Copyright TechTarget - All rights reserved https://cyber.law.harvard.edu/rss/rss.html Techtarget Feed Generator en Mon, 24 Aug 2026 18:50:52 GMT https://www.techtarget.com/it-infrastructure editor@techtarget.com <p>A high-performance GPU cluster is only as effective as the network infrastructure that supports it.</p> <p>Traditional enterprise networks connect users, applications, databases, cloud services and the internet. Organizations know how to transmit these workloads by understanding -- among many other attributes -- how many servers are required to exchange data at sustained rates and with low latency.</p> <p>AI changes this relationship between networking and compute. <a href="https://www.techtarget.com/ai/tip/How-to-train-an-LLM-on-your-own-data"></a></p> <p><a href="https://www.techtarget.com/ai/tip/How-to-train-an-LLM-on-your-own-data">Training a large AI model</a> distributes processing across hundreds or thousands of GPUs that continuously exchange gradients, model parameters and other intermediate data.</p> <p>This makes network performance critical to workload efficiency. Enterprises planning AI-ready infrastructure must therefore determine where <a href="https://www.techtarget.com/it-infrastructure/tip/How-to-ensure-network-performance-and-reliability">high-performance networking</a> is necessary, how much capacity their workloads require and if additional network investment can improve accelerator utilization -- the amount of time GPUs and other related hardware actively process data.</p> <section class="section main-article-chapter" data-menu-title="Different traffic patterns with AI workloads"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Different traffic patterns with AI workloads</h2> <p>AI workloads do not have a single traffic profile. Training, fine-tuning and inference use accelerators differently, which changes how much data moves across the network and how sensitive the workload is to bandwidth, latency and congestion.</p> <p>A traffic pattern depends on how the model is divided across the compute cluster. In data parallelism, each GPU works on a different portion of the training data and then synchronizes its updates with the others. Tensor, pipeline and expert parallelism split the model or its computation across multiple GPUs instead; these techniques require frequent communication among them.</p> <p>This makes training traffic highly synchronized. Google has <a target="_blank" href="https://cloud.google.com/blog/products/networking/introducing-virgo-megascale-data-center-fabric" rel="noopener">observed</a> sub-millisecond, line-rate bursts in AI training workloads as GPUs finish compute phases and begin exchanging data across the network.</p> <p>Even if one network path is delayed, the other accelerators involved might have to wait before the next stage of computation can begin. For this reason, training depends on more than raw link speed.</p> <p>A network might provide enough average bandwidth but still experience queue buildup or congestion when many accelerators transmit simultaneously. Predictable latency and the ability to sustain large east-west transfers become important metrics.</p> <p>Training also creates substantial traffic as datasets and model states move between storage systems and the compute cluster. Network teams need to plan for <a href="https://medium.com/@nikheelvs/gpu-to-gpu-communication-unlocking-parallelism-beyond-the-core-a80de2974078">accelerator-to-accelerator communication</a> alongside storage and orchestration traffic. Fine-tuning models produces many of the same traffic patterns but at a much wider range of scales.</p> <p>Inference, where the AI model applies what it's learned to generate results, is more sensitive to how fast applications can send data to a model and receive the answer. For example, real-time applications might depend on consistently low latency, while high-volume inference services need sufficient throughput to handle simultaneous requests. In generative AI, the additional metrics, such as time to first token and tail latency, can also become important.</p> <p>Inference is also becoming less of a north-south traffic generator as larger models are distributed across multiple GPUs or servers. <a href="https://www.techtarget.com/ai/definition/retrieval-augmented-generation">Retrieval-augmented generation</a> can add further traffic among inference services, vector databases, storage systems and enterprise applications.</p> </section> <section class="section main-article-chapter" data-menu-title="The shift from north-south to east-west traffic"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The shift from north-south to east-west traffic</h2> <p>North-south traffic enters or leaves the data center as employees access applications, customers connect to services, or systems communicate with the internet and cloud. But in <a href="https://www.techtarget.com/ai/feature/GenAI-data-center-infrastructure-reshapes-business-processes">AI data centers,</a> east-west traffic remains within the data center or cloud environment where the data moves among servers, GPUs, storage systems and other infrastructure services.</p> <p>This traffic pattern places pressure on traditional hierarchical network designs. For example, a three-tier network connects servers through access, aggregation and core layers. These designs can use oversubscription because the bandwidth available to servers exceeds the uplink capacity.</p> <p>With synchronized AI traffic, however, many servers attempt to use those uplinks together. AI clusters, therefore, adopt flatter leaf-spine (or <a href="https://www.techtarget.com/it-infrastructure/definition/Clos-network">Clos</a>) fabrics with lower oversubscription. In this design, servers connect to leaf switches, while each leaf connects across a set of spine switches.</p> <p>This creates multiple paths between endpoints and reduces the number of network tiers traffic must cross. Some large AI fabrics go even further toward non-blocking designs, where the network has sufficient capacity to support a much larger simultaneous allocation of traffic injected by the connected systems.</p> <p>Google's Virgo Network architecture, for example, uses a flat two-layer, non-blocking topology with high-radix switches to provide scale-out bandwidth for accelerators. This makes bisection bandwidth increasingly important. It describes how much traffic the fabric can carry simultaneously between two network segments.</p> <p>Therefore, network teams must consider how many accelerators can communicate simultaneously, how much traffic they inject into the fabric and how much contention the workload can tolerate.</p> </section> <section class="section main-article-chapter" data-menu-title="Lossless networking and GPU performance dependencies"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Lossless networking and GPU performance dependencies</h2> <p>When congestion causes packets to be dropped, higher-layer protocols such as TCP can retransmit the missing data and adjust the sending rate. For many enterprise applications, an occasional retransmission increases latency.</p> <p>But this step can generate a substantial consequence in a distributed AI cluster. If congestion delays one participant, the other GPUs involved in that operation may have to wait. To that end, low-loss or lossless networking is a must-have for <a href="https://www.techtarget.com/searchenterpriseai/tip/Best-practices-for-building-scalable-AI-infrastructure">enterprise AI workloads</a> that are tightly connected across accelerators.</p> <p>The goal of a lossless network is not to eliminate congestion, but to prevent it from causing packet drops. This requires coordinated buffer management, congestion control and flow control.</p> <p>Operating such networks requires comprehensive visibility. For example, a 400 GbE link that appears underutilized could still experience short bursts that fill queues and disrupt a training operation.</p> </section> <section class="section main-article-chapter" data-menu-title="How enterprises should plan an AI-ready network"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How enterprises should plan an AI-ready network</h2> <p>The following steps can help organizations build an AI-ready network.</p> <ol class="default-list"> <li>Begin with the workload. A large, distributed training cluster, a fine-tuning environment and an enterprise inference service each places different demands on the network.</li> <li>Determine how many accelerators need to communicate, where models and data will stay, if inference is centralized or distributed and how fast those requirements will scale.</li> <li>Understand where that traffic will travel. Map the data paths among compute, storage, applications and external services. This will determine if the organization needs separate networking domains.</li> <li>Capacity planning should focus on where bandwidth is required. Consider the connectivity available to each server, the combined rate at which nodes can inject traffic storage, throughput, uplink capacity, oversubscription and redundancy.</li> <li>Headroom is important because AI traffic <a href="https://www.techtarget.com/it-infrastructure/tip/Checking-for-high-network-utilization-on-a-switched-network">can be bursty</a>. The appropriate oversubscription level for the link utilization depends on how much contention a particular workload can tolerate.</li> <li>Decide how much operational complexity you are prepared to support. High-performance Ethernet fabrics using Remote Direct Memory Access over Converged Ethernet requires detailed congestion management, telemetry and configuration. The choice depends on the available networking skills, tools and operational processes.</li> <li>Observability plays a strong role, monitoring everything from queue behavior, packet drops and explicit congestion notification and priority flow control events to latency and congestion hotspots. This helps network teams correlate network behavior with GPU utilization.</li> </ol> <ol type="1" start="3" class="default-list"></ol> <p>Comprehensive planning treats network spending in relation to the value of the compute it supports. The right networking for AI workloads architecture is the one that provides enough bandwidth, latency consistency and resilience for the workload without affecting performance.</p> <p><i>Abhishek Jadhav is a technology journalist covering AI infrastructure, semiconductors and advanced computing systems.</i></p> </section> Distributed AI workloads reshape the architectures and strategies that enterprises traditionally relied on to plan network work capacity, manage congestion and support east-west traffic. https://cdn.ttgtmedia.com/rms/onlineimages/iot_g1226985345.jpg https://www.techtarget.com/it-infrastructure/tip/How-AI-workloads-affect-enterprise-network-design Mon, 24 Aug 2026 16:15:00 GMT <p>Enterprise cloud strategy has reached a turning point with multi-cloud deployments. What was once a technical architecture decision is now a core business and financial consideration. Relying on a single cloud provider concentrates on operational risk, limits negotiating leverage and constrains how quickly organizations can adapt to changing technology, market and regulatory demands.</p> <p><a href="https://www.techtarget.com/searchcloudcomputing/definition/multi-cloud-strategy">Multi-cloud deployment models</a> distribute workloads across cloud providers, offering these significant advantages:</p> <ul class="default-list"> <li>Improved resilience.</li> <li>Infrastructure aligned with <a href="https://www.techtarget.com/it-infrastructure/tip/Privacy-compliance-and-governance-are-changing-development">regional compliance requirements</a>.</li> <li>Access to best-in-class capabilities that accelerate innovation.</li> </ul> <p>While multi-cloud adds operational complexity, it also enables disciplined cost management through competitive pricing and <a href="https://www.techtarget.com/searchcloudcomputing/tip/A-month-by-month-roadmap-for-the-three-phases-of-FinOps">FinOps practices</a>.</p> <p>For IT leaders, the question is no longer whether to adopt multi-cloud, but rather how to implement it in a way that balances cost, risk, performance and governance while positioning the organization for long-term growth and agility.</p> <p>This article outlines the risks of single-vendor cloud strategies before exploring financial ramifications, operating models and a practical execution roadmap.</p> <section class="section main-article-chapter" data-menu-title="The cost of standing still: Single-cloud strategy risks"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The cost of standing still: Single-cloud strategy risks</h2> <p>A single-cloud strategy creates concentrated exposure across financial, operational and strategic dimensions. Single-cloud environments are often the hidden liability on the balance sheet. They can also limit business continuity options and compromise compliance management.</p> <p>Single-cloud deployment models surface the following risks:</p> <ul class="default-list"> <li><b>Vendor concentration.</b> Using a single cloud can create a risk and pricing power imbalance due to vendor lock-in and monopolization of the organization's data and services. Organizations are also vulnerable to shifting service priorities by the vendor.</li> <li><b>Financial exposure.</b> Outages affecting the vendor can result in lost revenue, service-level agreement penalties and reputational damage.</li> <li><b>Regulatory and data sovereignty challenges.</b> <a href="https://www.techtarget.com/it-strategy/tip/Why-CIOs-must-prepare-for-geopolitical-disruption">Global operations face increasing scrutiny</a> from regulatory bodies that many cloud vendors cannot satisfy.</li> <li><b>Innovation constraints.</b> A single vendor's ecosystem does not offer the specialization, flexibility and agility needed for business innovation.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="The financial case for multi-cloud"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The financial case for multi-cloud</h2> <p>The financial case for multi-cloud extends beyond simple cost comparison; it's about optimizing total economic value while reducing risk. These numbers are more difficult to extract but clearly demonstrate the financial benefits of multi-cloud deployments.</p> <p><b>Total cost of ownership.</b> A multi-cloud strategy can introduce incremental complexity and operational overhead. However, these costs are often offset, as organizations can avoid the hidden premiums of single-vendor dependencies, including escalating <a href="https://www.computerweekly.com/feature/Cloud-egress-costs-What-they-are-and-how-to-dodge-them">data egress fees</a>, limited pricing leverage and constrained architectural choices.</p> <p><b>Return on investment.</b> ROI is driven less by immediate savings and more by risk-adjusted outcomes. Cross-cloud residency mitigates downtime, which helps organizations avoid significant revenue loss and reputational damage. Selecting the best services -- regardless of vendor -- for a business requirement enables faster innovation, directly supporting revenue growth and competitive differentiation.</p> <p><b>FinOps.</b> <a href="https://www.techtarget.com/searchcloudcomputing/tip/The-business-case-for-FinOps-to-control-cloud-spending">Disciplined FinOps</a> is essential for realizing these benefits. Multi-cloud environments require centralized cost visibility, real-time usage tracking and clear accountability across business units. Cloud spend is no longer a fixed expense. Instead, organizations optimize workloads, rebalance provider usage and align infrastructure investments with business priorities.</p> </section> <section class="section main-article-chapter" data-menu-title="Balancing risk, performance and compliance"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Balancing risk, performance and compliance</h2> <p>Multi-cloud environments balance risk, performance and compliance. Realizing the financial and business continuity benefits of a multi-cloud strategy requires ongoing fine-tuning.</p> <p><b>Risk distribution.</b> Multi-cloud distributes security and availability risks across a broader surface, reducing the <a href="https://www.techtarget.com/it-strategy/feature/AWS-cloud-outage-reveals-vendor-concentration-risk">impact of a vendor-level incident</a> on the organization's operations. This helps organizations avoid single points of failure, and it enables scenario planning for regional outages.</p> <p><b>Performance tradeoffs.</b> Synchronizing data and service between vendors and regions introduces potential performance issues, including <a href="https://www.techtarget.com/searchdatacenter/feature/Data-gravity-and-its-role-in-data-center-efficiency">data gravity</a> and latency challenges. Proprietary data structures must also be addressed. However, multi-cloud can improve performance by placing resources nearer to consumers.</p> <p><b>Compliance.</b> Multi-cloud serves as a compliance enabler by aligning workloads with regional and industry regulations, helping organizations avoid penalties. <a href="https://www.techtarget.com/searchcloudcomputing/tip/A-data-sovereignty-primer-for-cloud-admins">Data residency and sovereignty</a> compliance are of particular importance.</p> <p>Multi-cloud deployments introduce new challenges while mitigating others. Strong design and governance are required for intentional tradeoffs.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Multi-cloud is no longer optional; it is a strategic lever for resilience, financial control and innovation. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> </section> <section class="section main-article-chapter" data-menu-title="What a multi-cloud deployment model looks like in practice"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What a multi-cloud deployment model looks like in practice</h2> <p>Three primary use cases drive a multi-cloud strategy. These choices inform which multi-cloud deployment model the organization will select:</p> <ul class="default-list"> <li>Disaster recovery and failover.</li> <li>Regulatory segmentation and compliance.</li> <li>Utilizing specialized services (AI, analytics, etc.).</li> </ul> <p>Consider the following common deployment models:</p> <ul class="default-list"> <li><b>Active-active across cloud vendors.</b> This model runs critical systems and application stacks in multiple clouds simultaneously via load balancers. The goal is near-zero recovery point and recovery time objectives and resilience to provider outages.</li> <li><b>Workload-specific placement.</b> This model places workloads on the cloud platform that best fits its needs. It typically involves a primary cloud provider hosting most services, with niche applications on specific vendor clouds.</li> <li><b>Hybrid, multi-cloud combinations.</b> This model combines on-premises capabilities with multiple public clouds. Regulated or latency-sensitive data is hosted on-premises.</li> </ul> <p>An organization's business requirements will drive the selection of the right model.</p> </section> <section class="section main-article-chapter" data-menu-title="Operating model: From strategy to execution"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Operating model: From strategy to execution</h2> <p>Shifting to a multi-cloud deployment requires resource, governance and workflow changes.</p> <ul class="default-list"> <li><b>Organizational changes.</b> <a href="https://www.techtarget.com/searchcloudcomputing/tip/How-to-build-a-cloud-center-of-excellence">Establish a cloud center of excellence</a> and platform engineering teams to oversee deployments and support.</li> <li><b>Skills and talent development.</b> Expertise across diverse cloud vendors and environments is required. Most cloud engineers and architects are <a href="https://www.techtarget.com/searchcloudcomputing/tip/Is-there-still-a-cloud-skills-gap">only skilled on one platform</a>.</li> <li><b>Vendor management.</b> Establish a vendor management team that maximizes each cloud provider's features and recognizes negotiation opportunities to optimize pricing.</li> <li><b>Standardization vs. flexibility tradeoffs.</b> Single-cloud deployments offer standardization, simplifying deployment and support. Balance this with the flexibility and resilience offered by multi-cloud environments.</li> </ul> <p>Multi-cloud success depends more on the operating model's maturity, thoughtfulness and accuracy than on the technology platforms where it resides.</p> </section> <section class="section main-article-chapter" data-menu-title="Key architectural concerns"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Key architectural concerns</h2> <p>Multi-cloud architects must design effective deployments that maximize benefits and minimize costs. Key factors include managing workloads, networking and access control:</p> <ul class="default-list"> <li><b>Workload placement.</b> Criteria include cost, latency and compliance across vendors.</li> <li><b>Intercloud networking.</b> Challenges include secure, high-performing networking across cloud providers to ensure availability and application support.</li> <li><b>Identity and access management.</b> Authentication and access controls across multi-cloud environments are more complex, requiring either <a href="https://www.techtarget.com/cybersecurity/tip/CISOs-guide-to-centralized-vs-federated-security-models">centralized or federated solutions</a><a name="_msoanchor_1" language="JavaScript" href="#_msocom_1" id="_anchor_1">[DG1]</a>&nbsp;<a></a>.</li> <li><b>Tooling.</b> Tooling must enable visibility across multiple cloud environments and integrate on-premises capabilities. Include observability, automation and cost management platforms.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Create an action plan"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Create an action plan</h2> <p>A successful multi-cloud strategy starts with a structured, business-aligned approach to execution. Use the following steps to structure the action plan:</p> <ol class="default-list"> <li><b>Assess current workloads for cloud suitability.</b> Evaluate risk, performance, cost sensitivity, regulatory requirements and interdependencies. Not all workloads justify being distributed. Prioritize those that benefit from resilience and flexibility.</li> <li><b>Design a hybrid/multi-cloud architecture with standardized governance.</b> Ensure consistency in security, compliance and cost controls. Establish clear policies and reference models that scale across cloud providers and internal business units. Define guardrails for security, cost and operations.</li> <li><b>Implement cross-cloud monitoring and automation tools.</b> <a href="https://www.techtarget.com/searchcloudcomputing/tip/What-is-multi-cloud-visibility-and-how-do-I-achieve-it">Provide unified visibility</a> into performance, spending and risk. Automation reduces operational overhead while improving reliability and response times.</li> <li><b>Define vendor selection criteria aligned with business priorities.</b> Balance cost, geographic coverage, innovation capabilities and compliance requirements.</li> <li><b>Use FinOps and risk management KPIs.</b> Track cost efficiency, uptime and vendor concentration to show the strategy delivers measurable business value over time.</li> </ol> </section> <section class="section main-article-chapter" data-menu-title="Measuring success: KPIs that matter"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Measuring success: KPIs that matter</h2> <p>Measurable business outcomes are essential for managing value over time. Use the following <a href="https://www.techtarget.com/searchcloudcomputing/tip/FinOps-KPIs-and-metrics-to-optimize-cloud-expenditures">KPIs and metrics</a> to evaluate the success of a multi-cloud strategy across financial performance, operations, and risk management.</p> <h3>Financial and efficiency metrics</h3> <ul class="default-list"> <li><b>Unit cost per workload.</b> Cost per transaction, user or application instance across clouds.</li> <li><b>Cost of data transfer (egress ratio).</b> Percentage of total spend tied to intercloud or outbound data movement.</li> <li><b>Resource utilization.</b> Percentage of provisioned vs. active compute/storage resources.</li> <li><b>Budget variance.</b> A measure of forecast spend vs. actual spend over time.</li> </ul> <h3>Operational and performance metrics</h3> <ul class="default-list"> <li><b>Deployment velocity and time to market.</b> Time taken to deploy capabilities from ideation to user availability.</li> <li><b>Cross-cloud failover time.</b> Time required to shift workloads between providers during disruption.</li> <li><b>Deployment frequency across environments.</b> How often applications are released across multiple clouds.</li> <li><b>Mean time to detect and respond.</b> Efficiency of incident detection and resolution.</li> </ul> <h3>Risk and resilience metrics</h3> <ul class="default-list"> <li><b>Uptime.</b> Measure of system and service availability.</li> <li><b>Workload portability score.</b> Percentage of applications that can be moved between clouds with minimal <a href="https://www.techtarget.com/searchCloudComputing/tip/Use-the-7-Rs-to-develop-an-app-modernization-strategy">rework</a>.</li> <li><b>Provider dependency ratio.</b> Share of critical workloads tied to a single provider.</li> </ul> <h3>Governance and compliance metrics</h3> <ul class="default-list"> <li><b>Policy compliance rate.</b> Percentage of resources adhering to defined security, cost and governance policies.</li> </ul> <p>These metrics offer leadership early warning signals pertaining to <a target="_blank" href="https://www.youtube.com/watch?v=Eq4NRgIkBuU" rel="noopener">cost drift</a>, risk concentration and operational inefficiencies before they negatively impact the business.</p> </section> <section class="section main-article-chapter" data-menu-title="Multi-cloud as a foundation for resilient growth"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Multi-cloud as a foundation for resilient growth</h2> <p>Multi-cloud is no longer optional; it is a strategic lever for resilience, financial control and innovation. Organizations that act now will reduce risk while gaining competitive flexibility. Begin by assessing current cloud exposure and defining a deliberate multi-cloud roadmap aligned to business priorities.</p> <p><em>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to TechTarget Editorial, The New Stack and CompTIA Blogs.</em></p> </section> With a single-cloud strategy, organizations risk vendor lock-in, financial exposure and regulatory hurdles. A multi-cloud deployment model is essential for today's IT operations. https://cdn.ttgtmedia.com/visuals/LeMagIT/hero_article/Cloud_sky.jpg https://www.techtarget.com/it-infrastructure/tip/Why-multi-cloud-should-be-your-default-deployment-model Mon, 24 Aug 2026 11:24:00 GMT <p>Distributed computing is not just an infrastructure concern -- it is a driver of revenue, resilience and experience, making it critical to an organization's technology strategy.</p> <p>Distributed architecture is a system design approach in which computing, data processing and services are deployed across multiple locations -- core, edge and cloud -- to optimize performance, resilience and proximity to users or data, rather than relying on a single centralized environment.</p> <p>Centralized resources, such as those found in standard cloud deployments, have significant limitations for real-time, data-intensive or location-sensitive use cases. This article helps IT leaders determine <a href="https://www.techtarget.com/searchdatacenter/opinion/Distributed-computing-The-infrastructure-shift-AI-demands">where distributed computing creates value</a>, how to manage distributed infrastructure and what governance practices are necessary.</p> <section class="section main-article-chapter" data-menu-title="Where distributed computing creates value"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Where distributed computing creates value</h2> <p>Distributed architecture makes sense when performance, resilience and real-time outcomes matter most. Common design triggers include:</p> <ul class="default-list"> <li>Ultra-low latency requirements such as real-time analytics and automation.</li> <li>Data gravity and locality such as regulatory, <a href="https://www.techtarget.com/searchSecurity/tip/Data-sovereignty-compliance-challenges-and-best-practices">sovereignty</a> or performance constraints.</li> <li>Bandwidth cost and efficiency considerations.</li> </ul> <p>Not all data requires a distributed infrastructure. In many cases, standard cloud platforms are sufficient. Centralized cloud still wins in areas such as batch processing and AI model training.</p> <p>To identify where distributed computing creates value, begin by mapping use cases to outcomes such as revenue growth, cost optimization and customer experience differentiation. Use a <a href="https://untools.co/decision-matrix/">decision matrix</a> to select among centralized, hybrid or distributed architectures.</p> <figure class="main-article-image full-col" data-img-fullsize="https://www.techtarget.com/rms/onlineimages/the_distributed_computing_process-f.png"> <img data-src="https://www.techtarget.com/rms/onlineimages/the_distributed_computing_process-f_mobile.png" class="lazy" data-srcset="https://www.techtarget.com/rms/onlineimages/the_distributed_computing_process-f_mobile.png 960w,https://www.techtarget.com/rms/onlineimages/the_distributed_computing_process-f.png 1280w" alt="distributed computing example" height="269" width="560"> <figcaption> <i class="icon pictures" data-icon="z"></i>An example showing how networks, servers and computers are structured in distributed computing. </figcaption> <div class="main-article-image-enlarge"> <i class="icon" data-icon="w"></i> </div> </figure> </section> <section class="section main-article-chapter" data-menu-title="Manage infrastructure across core, cloud and edge"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Manage infrastructure across core, cloud and edge</h2> <p>Use the following approaches to translate strategy into operational reality.</p> <h3>Investment prioritization</h3> <p>Investment decisions must be driven by clearly defined business outcomes, not just aspirations to <a href="https://www.techtarget.com/searchcloudcomputing/tip/A-primer-on-modernization-strategies-for-legacy-systems">modernize architecture</a>. Prioritize distributed deployments that directly support revenue-generating digital services, latency-sensitive operations or regulatory mandates. Evaluate each investment against its expected business impact to ensure resources are allocated effectively. Follow these four steps;</p> <ol class="default-list"> <li>Align investments with high-value use cases first.</li> <li>Plan for incremental distributed deployments.</li> <li>Avoid broad rollouts and over-distribution, such as ungoverned edge sprawl without ROI.</li> <li>Balance Capex and Opex across environments.</li> </ol> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Select investments for high impact and scalability, balancing capital-intensive edge infrastructure with flexible cloud services. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Select investments for high impact and scalability, balancing capital-intensive edge infrastructure with flexible cloud services.</p> <h3>Architectural patterns</h3> <p>Distributed architecture needs intentional workload placement and modern design patterns. Aligning use cases creates a cohesive system rather than a fragmented infrastructure. Several common architectural patterns fit specific use cases, which can include:</p> <ul class="default-list"> <li>Hybrid and multi-cloud integration enable flexibility.</li> <li>Containerization and orchestration help standardize deployments across core, edge and cloud environments.</li> <li>Place latency-sensitive processing closer to the edge.</li> <li>Place aggregation and analytics in centralized environments.</li> </ul> <h3>Operational implications</h3> <p>Distributed architectures add significant complexity compared to centralized models. Teams must manage diverse infrastructure, differing connectivity capabilities and geographically dispersed assets. Manual operations do not scale in distributed environments, so plan accordingly.</p> <ul class="default-list"> <li>Expect increased operational complexity over centralized models.</li> <li><a href="https://www.techtarget.com/searchcloudcomputing/infographic/By-the-numbers-How-upskilling-fills-the-IT-skills-gap">Address skills gaps</a> and retain essential talent.</li> <li>Unify observability, orchestration and remote lifecycle management.</li> </ul> <p>Centralized control planes and automation provides reliability, as well as enforces consistency and performance across all sites.</p> </section> <section class="section main-article-chapter" data-menu-title="Governance in distributed environments"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Governance in distributed environments</h2> <p>Effective governance addresses risks, maintains control and ensures consistency at scale.</p> <ul class="default-list"> <li><b>Lifecycle management.</b> Manage provisioning, updating and retiring distributed assets.</li> <li><b>Security enforcement.</b> <a href="https://www.techtarget.com/cybersecurity/tip/Perimeter-to-posture-A-roadmap-to-zero-trust-maturity">Implement zero-trust</a> models across environments and recognize the expanded attack surface at the edge.</li> <li><b>Data consistency.</b> Resolve synchronization challenges across nodes to balance consistency, availability and latency.</li> <li><b>Regulatory and compliance.</b> Maintain data sovereignty obligations and auditability.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Talent and organizational readiness"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Talent and organizational readiness</h2> <p>Shifting to a distributed architecture is an organizational long-term change rather than a technical deployment project. Success depends on teams designing, operating and optimizing systems that span core, cloud and edge environments. This requires new <a href="https://www.techtarget.com/searchCloudComputing/tip/How-to-use-AI-to-bridge-the-cloud-skills-gap">skill sets</a> across engineering, data streaming, security and operations. These talents are often scarce, requiring development within existing talent pools.</p> <p>Traditional silos between infrastructure, applications and operations teams impede distributed environments. Many organizations use platform-based models, where centralized teams provide standardized tooling, automation and guardrails, while product teams retain the necessary flexibility in how they deploy and run workloads. DevOps and site reliability engineering are critical to managing complexity at scale.</p> <p>Failure to account for these challenges means the distributed architecture could become fragmented, costly and difficult to sustain.</p> <p><iframe title="Key metrics and KPIs to track" aria-label="Table" id="datawrapper-chart-OFwnr" src="https://datawrapper.dwcdn.net/OFwnr/1/" scrolling="no" frameborder="0" style="width: 0; min-width: 100% !important; border: none;" height="595" data-external="1"></iframe></p> <p> <script type="text/javascript">(function(){function e(){window.addEventListener(`message`,function(e){if(e.data[`datawrapper-height`]!==void 0){var t=document.querySelectorAll(`iframe`);for(var n in e.data[`datawrapper-height`])for(var r=0,i;i=t[r];r++)if(i.contentWindow===e.source){var a=e.data[`datawrapper-height`][n]+`px`;i.style.height=a}}})}e()})();</script> </p> </section> <section class="section main-article-chapter" data-menu-title="Linking architecture to business outcomes"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Linking architecture to business outcomes</h2> <p>Distributed architectures do not address every operational requirement, nor do they support every workload. The essential element is determining whether distribution offers a measurable impact on essential business processes.</p> <p>Distributed computing is a continuous capability that scales with evolving business needs, not a one-time project. It enables:</p> <ul class="default-list"> <li>New digital innovations and products.</li> <li>Real-time decision-making.</li> <li>Enhanced customer experiences.</li> </ul> <p>Centralized designs have become a constraint -- reevaluate where the organization runs, not just what it runs. If the architecture can't support real-time or location-aware experiences, a competitor's will.</p> </section> Now is the time to align infrastructure decisions with the speed and expectations of your customers. See how distributed computing can create value and what to expect. https://cdn.ttgtmedia.com/rms/onlineimages/iot_g1224942277.jpg https://www.techtarget.com/it-infrastructure/tip/How-distributed-computing-drives-revenue-and-resilience-in-cloud Thu, 20 Aug 2026 10:48:00 GMT <p>In June 2026, HTTP gained its first new general-purpose method in over a decade – QUERY.</p> <p>Organizations increasingly need to submit large or complex read/query criteria. GET pushes that information into URLs, while POST carries it in a body without explicitly safe or idempotent semantics.</p> <p>As of June 2026, the HTTP QUERY method is a standard as defined in <a target="_blank" href="https://www.rfc-editor.org/info/rfc10008/" rel="noopener">RFC 10008</a>. It is a safe, <a href="https://www.theserverside.com/tip/Idempotent-HTTP-methods-and-REST">idempotent method</a> that carries query content in the request body.</p> <p>For enterprises, this means greater reliability for repeatable queries, better separation of query intent from mutation and stronger foundations for logging, governance and compliance.</p> <p>This article outlines the limitations of <a href="https://www.theserverside.com/video/GET-vs-POST-Whats-the-difference">GET and POST</a> before introducing the new QUERY method, noting its advantages and offering adoption considerations. Organizations do not need to update current APIs; QUERY is a targeted option for specific workloads.</p> <section class="section main-article-chapter" data-menu-title="What's wrong with GET and POST?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What's wrong with GET and POST?</h2> <p>Traditional methods for managing information have significant problems for today's workloads. Enterprises commonly use these methods to retrieve data.</p> <ul class="default-list"> <li><b>GET.</b> Simple, read-oriented queries with parameters in the URL.</li> <li><b>POST.</b> Complex queries whose inputs are too large or cumbersome for a URL.</li> </ul> <p>However, these methods have operational consequences. Long or highly encoded URLs can encounter size and interoperability constraints across proxies, gateways and other intermediaries. Furthermore, query information in URLs is more likely to surface in logs, browser history, bookmarks and other systems. In particular, using POST for read-only queries obscures the intended semantics and can limit the behavior of caches and automated retry mechanisms.</p> <p>These consequences go beyond protocol design. They affect service reliability, incident recovery, data exposure, auditability and operational cost.</p> </section> <section class="section main-article-chapter" data-menu-title="What is the new HTTP QUERY method?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What is the new HTTP QUERY method?</h2> <p>As defined in RFC 10008: The HTTP QUERY Method, a QUERY is a request that asks a target resource to perform a query using content in the request body in a safe and idempotent way. In this context, safe and idempotent have specific meanings:</p> <ul class="default-list"> <li><b>Safe.</b> The request is intended for retrieval/querying, not changing server state.</li> <li><b>Idempotent.</b> Repeating the same request does not create additional state changes, making automated repetition safer.</li> </ul> <p>The key distinction is that QUERY combines the body-based input of POST with the safe, idempotent semantics of GET. QUERY also supports caching.</p> </section> <section class="section main-article-chapter" data-menu-title="QUERY vs GET and POST: The strategic difference"> <h2 class="section-title"><i class="icon" data-icon="1"></i>QUERY vs GET and POST: The strategic difference</h2> <p>The strategic benefit isn't that QUERY makes APIs inherently faster or more secure. It gives organizations more precise control for complex read operations.</p> <p>GET, POST and QUERY serve different purposes. GET is best for straightforward retrieval, QUERY for complex read-only queries and POST for when operations might create or change state. The distinction carries beyond API design to offer three practical advantages:</p> <ul class="default-list"> <li><b>Reliability.</b> Due to its idempotent nature, QUERY provides a clearer basis for retry and recovery strategies when networks, gateways or downstream services fail.</li> <li><b>Observability.</b> The dedicated QUERY method is differentiated in log files containing POST results, providing clarity for monitoring, classification and <a href="https://www.computerweekly.com/blog/CW-Developer-Network/SmartBear-fills-honeypot-of-AI-ready-API-governance">API governance</a>.</li> <li><b>Compliance and governance.</b> QUERY carries search criteria in the request body, reducing unnecessary exposure of information.</li> </ul> <p>Note that implementing QUERY is a targeted <a href="https://www.techtarget.com/it-infrastructure/tip/A-primer-on-modernization-strategies-for-legacy-systems">modernization</a> task, not an enterprise-wide migration project. It gives complex read operations their own explicit method, potentially reducing operational ambiguity and risk without requiring organizations to replace APIs that already work well.</p> </section> <section class="section main-article-chapter" data-menu-title="Developer use cases"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Developer use cases</h2> <p>Developers and engineers can get the most value from the QUERY method in specific use cases, such as:</p> <ul class="default-list"> <li>Complex search or filter APIs.</li> <li>Analytics and reporting queries.</li> <li>APIs accepting structured query languages.</li> <li>Requests that routinely approach URI-size or encoding constraints.</li> </ul> <p>Note that API gateways, WAFs, proxies, SDKs, <a href="https://www.techtarget.com/it-infrastructure/tip/End-to-end-network-observability-for-AI-workloads">observability platforms</a> and application frameworks must recognize and correctly pass the new QUERY method. Watch for vendor updates.</p> </section> <section class="section main-article-chapter" data-menu-title="Deployment options"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Deployment options</h2> <p>Using QUERY entails specific advantages and trade-offs. For most organizations, integration is probably worthwhile for future applications. Some companies will want to retool existing APIs to add QUERY.</p> <p><b>Option A: Retain existing GET/POST use</b></p> <ul class="default-list"> <li>Advantage: Lowest immediate friction.</li> <li>Trade-off: Appropriate when current APIs are reliable and query inputs are simple.</li> </ul> <p><b>Option B: Introduce QUERY selectively</b></p> <ul class="default-list"> <li>Advantage: Best balance of risk and modernization.</li> <li>Trade-off: Add QUERY for new or high-value complex-query endpoints while retaining existing configurations.</li> </ul> <p><b>Option C: Broad API migration (probably unnecessary for most organizations)</b></p> <ul class="default-list"> <li>Advantages: Potential long-term consistency.</li> <li>Trade-offs: Possible higher compatibility, testing and infrastructure risk.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Turn QUERY into a targeted risk-reduction initiative"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Turn QUERY into a targeted risk-reduction initiative</h2> <p>QUERY fills a specific gap between GET and POST, so its use is situational. IT leaders should assess QUERY-ready use cases during the next API governance cycle. The <a href="https://www.techtarget.com/it-infrastructure/tip/Modern-software-quality-metrics-that-matter">results of that assessment</a> will guide the next steps.</p> <ul class="default-list"> <li>Begin with selective adoption, not a wholesale rewrite.</li> <li>Create an API-method policy defining when teams should use GET, QUERY and POST.</li> <li>Identify a small number of complex read/query endpoints as pilots.</li> <li>Confirm compatibility with gateway, WAF, proxy, logging, monitoring, caching and security tools.</li> <li>Measure retry success, latency, error rates, observability coverage and sensitive data exposure.</li> </ul> <p>Start today with one high-level question: Which existing APIs would be safer, more observable and easier to operate if complex queries had an explicit, idempotent HTTP method?</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides and contributes extensively to Informa TechTarget, The New Stack and CompTIA Blogs.</i></p> </section> HTTP QUERY gives complex read operations a safe, idempotent alternative to GET and POST. Learn its benefits, use cases and enterprise adoption options. https://cdn.ttgtmedia.com/rms/onlineimages/container_g1074391400.jpg https://www.techtarget.com/it-infrastructure/tip/A-look-at-HTTPs-new-QUERY-method Thu, 20 Aug 2026 10:00:00 GMT <p><a href="https://www.techtarget.com/searchdatacenter/definition/IT">IT</a> is all about data, while <a href="https://www.techtarget.com/whatis/definition/operational-technology">operational technology</a> focuses on the physical devices that control industrial operations and processes. Although they can run independently of each other, there are real benefits to IT/OT convergence when each side shares its strengths, including cost and security controls.</p> <p>The convergence of IT and OT is largely due to the growing adoption of <a href="https://www.techtarget.com/iotagenda/definition/Internet-of-Things-IoT">IoT</a> infrastructure and how IoT merges the processes and data types that OT oversees with those of traditional IT. IoT's allure is the potential for greater efficiencies, insights and monetization opportunities that merging devices, data and people into a single environment can engender.</p> <p>But IoT isn't the only compelling reason for IT/OT convergence. AI is pervasive in contemporary data centers and related computing resources, and it ups the IoT ante when it comes to integrating data sources, data processing and applications that can easily span multiple IT and OT environments within a single organization.</p> <section class="section main-article-chapter" data-menu-title="IT vs. OT: Key differences"> <h2 class="section-title"><i class="icon" data-icon="1"></i>IT vs. OT: Key differences</h2> <p>IT and OT are both network-based technical structures that link hundreds or thousands of pieces of equipment, but beyond that basic topology, there are more differences than similarities.</p> <p>Even the networks bear distinguishing differences. IT networks typically run atop a handful of standardized OSes, including Windows and Linux. <a href="https://www.techtarget.com/iotagenda/definition/Industrial-Internet-of-Things-IIoT">Industrial IoT</a> environments supporting OT might run on IT network OSes, but several proprietary OSes are more role-based and often tailored to specific industries or industrial processes. In some cases, companies will modify an off-the-shelf OS to develop one that fits their unique needs.</p> <p>The communication protocols that IT and OT infrastructures use can vary as well. Most IT networks are Ethernet-based, whether connected by cable -- copper or optical -- or wireless. All or part of an OT network can also use Ethernet as its protocol. However, because a single IoT implementation can potentially cover a much wider geographic area than an IT network, other protocols are used, such as LTE -- mostly 4G, with 5G adoption growing -- and low-power WAN communications in several implementations, including <a href="https://www.techtarget.com/whatis/definition/narrowband-IoT-NB-IoT">narrowband IoT</a> and LoRa.</p> <p>Those non-Ethernet protocols and carriers are often used to connect the remote devices to edge servers and edge storage before tapping into the organization's Ethernet infrastructure or a cloud service to store the collected data in a more centralized location.</p> <p>Although both IT and OT networks are conduits for data transfer, the size of the data packets and the speed at which they're transmitted, analyzed and used differ. IT systems are based on block-, file- or object-oriented file systems, which means they must support acceptable access performance for a variety of data types that can vary significantly in size and format. The data transmitted over an OT network might be extremely small -- even just a few bits at a time that easily describe a piece of gear's current state.</p> <p><iframe title="" aria-label="Table" id="datawrapper-chart-BxfKt" src="https://datawrapper.dwcdn.net/BxfKt/2/" scrolling="no" frameborder="0" style="width: 0; min-width: 100% !important; border: none;" height="838" data-external="1"></iframe></p> <p> <script type="text/javascript">(function(){function e(){window.addEventListener(`message`,function(e){if(e.data[`datawrapper-height`]!==void 0){var t=document.querySelectorAll(`iframe`);for(var n in e.data[`datawrapper-height`])for(var r=0,i;i=t[r];r++)if(i.contentWindow===e.source){var a=e.data[`datawrapper-height`][n]+`px`;i.style.height=a}}})}e()})();</script> </p> <p>OT is more device-focused than IT and uses data in real time to monitor and control physical devices -- in some cases, exercising that control instantaneously to ensure that processes are running correctly without interruption, and that worker safety systems aren't compromised.</p> <p>IT is user- and data-centric and often <a href="https://www.techtarget.com/searchbusinessanalytics/feature/8-types-of-bias-in-data-analysis-and-how-to-avoid-them">uses historical data for analyses</a> related to customer support, back-office reporting and marketing. Generally, IT admins are more attentive to security risks that could jeopardize data than to their users' physical well-being.</p> </section> <section class="section main-article-chapter" data-menu-title="What is IT?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What is IT?</h2> <p>IT is certainly the more recognizable of these two technologies and represents the critical infrastructure required for data processing. It's also arguably the more mature and advanced of the twin techs.</p> <p>IT systems are data-oriented, serving as repositories for corporate information and making that data available to business applications and the people who use them. IT's role is broad and extremely diverse, spanning systems that control and track accounting activities, sales and marketing, customer support, payroll processing and HR management.</p> <p>Physically, IT comprises the familiar components of computing systems, including servers, storage systems, network equipment and end-user devices. Most IT implementations are based on Ethernet network topologies with TCP/IP used for external data transmissions, principally using the internet or leased lines.</p> <p>Over the past couple of decades, the definition of IT has expanded to include cloud-based services and mobile computing devices, giving IT both a local and a remote presence. Internet connectivity is the primary enabler of IT's remote and cloud operations and a primary security concern.</p> </section> <section class="section main-article-chapter" data-menu-title="What is OT?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What is OT?</h2> <p>OT has been around for a long time, too, but it only achieved its own identity as automation was introduced into manufacturing and industrial systems, along with the need to network automated devices to gain appropriate control over factory-floor operations.</p> <p>Today, OT refers to the network of devices and software that's used in industrial, manufacturing and process control systems. The types of <a href="https://www.techtarget.com/iotagenda/feature/Determine-which-of-4-IoT-wireless-networks-fit-your-use-case">devices that hang off IoT networks</a> run the gamut from sensors, relays and other single-purpose circuitry on shop floors to end-users' laptops and smartphones. However, OT supports specialized gear in industrial environments that capture and relay data, enabling industrial equipment to perform specific tasks. Typically, these IoT installations are referred to as industrial control systems.</p> <p>Management for an ICS is often provided by <a href="https://www.techtarget.com/whatis/definition/SCADA-supervisory-control-and-data-acquisition">supervisory control and data acquisition</a> software, which handles some data gathering and processing, as well as equipment monitoring.</p> <p>Other OT systems include:</p> <ul class="default-list"> <li><b>Programmable Logic Controllers.</b> PLCs are ruggedized computers intended for industrial environments.</li> <li><b>Distributed Control Systems. </b>A DCS is a distributed, multi-node hardware and software system designed to control processes in an industrial setting.</li> <li><b>Manufacturing Execution Systems. </b>An MES is a software product that monitors manufacturing systems in real time.</li> </ul> <p>Unlike IT, which tends to turn over gear and update firmware frequently, OT devices can be deployed and left to operate for many years. If they are doing their jobs properly, they might not be updated regularly. That scenario often leads to situations where OT must manage several versions of a sensor OS or application software, further complicating the already tough task of managing hundreds or thousands of endpoints.</p> <p>One of the key challenges of OT is managing the sheer number of remote devices, making it essential to decentralize certain processes to avoid data center bottlenecks.</p> <p>The data gathered by OT systems -- particularly in IoT environments -- might serve dual functions. Its primary role is to ensure that the equipment being monitored by sensors continues to operate safely and efficiently. But OT networks might also need to transmit operational data back to a central site -- a data center or a cloud computing service -- for additional analysis alongside other data already collected by IT systems.</p> </section> <section class="section main-article-chapter" data-menu-title="IT vs. OT: Examples and use cases"> <h2 class="section-title"><i class="icon" data-icon="1"></i>IT vs. OT: Examples and use cases</h2> <p>As noted, IT is focused on gathering, storing and processing data so that it can be put into forms that are meaningful to human beings and provide some business or other value. IT systems are familiar to everyone, as they encompass computing resources for a broad spectrum of environments:</p> <ul class="default-list"> <li><b>Businesses.</b> From small mom-and-pop companies to international enterprises, business runs on data processed by computing systems, whether those systems are on a desktop or housed in a <a href="https://www.techtarget.com/searchdatacenter/tip/A-primer-on-hyperscale-data-centers">large data center</a>.</li> <li><b>Institutions.</b> Government entities of all sizes rely on IT to maintain civil operations and to serve their constituents.</li> <li><b>Entertainment and education.</b> IT reaches a vast number of homes using the internet to deliver news and related information, retail and other services, and streamed entertainment content.</li> </ul> <p>Some of the IT applications provided in the above scenarios include the following:</p> <ul class="default-list"> <li><b>Business systems.</b> Accounting, ERP, HR and employee development programs, CXM/CRM, end-user productivity apps, etc.</li> <li><b>Institutional applications. </b>Social Security Administration functions, local government websites/apps, Department of Motor Vehicles applications, IRS websites and services, etc.</li> <li><b>Entertainment and education. </b>Netflix and many other entertainment streaming services, video gaming, remote instruction, etc.</li> </ul> <p>OT, on the other hand, is much more focused on physical devices and their safe, continuous operation, with much of OT encompassing systems internal to particular machines or processes. OT is often an exercise in managing diverse, proprietary equipment.</p> <p>OT generally includes the following:</p> <ul class="default-list"> <li><b>Operation monitoring.</b> Applications and special hardware -- including proprietary networks -- are used to monitor shop-floor machinery and to correct anomalous operations in real time.</li> <li><b>Equipment maintenance.</b> Based on real-time data collection and analysis, many OT-based systems are used to determine when and where maintenance is required</li> <li><b>Remote management.</b> OT systems are also instrumental in ensuring that remote devices, such as computing endpoints and sensors, operate properly and send valid data back to a central management facility.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="What is IT/OT convergence?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What is IT/OT convergence?</h2> <p>Convergence isn't only about merging distinct IT and OT networks and normalizing disparate data; it's also about sharing the data and <a href="https://www.techtarget.com/searchsecurity/feature/Best-practices-to-secure-data-at-rest-in-use-and-in-motion">strengthening security across all data sources</a>, networks and computing resources. Much of the machine- and process-related data that OT systems collect can be useful to the external-facing side of the business for forecasting, planning, supply chain control and other decision-making processes. Conversely, the OT environment can use IT-hosted business data to adjust production systems for greater efficiency.</p> <p>A converged system incorporates multiple data paths, including corporate server-based networks and edge devices to collection points that can be central or satellite data centers or cloud services. Understanding the nature of the data being transmitted and its purpose is also critical to securing a converged network environment. Although IT might focus on securing data and detecting attempts to infiltrate the network, the OT security staff might be more concerned with anomaly detection to determine whether endpoint devices have been compromised by spurious data that could cause them to function improperly or unsafely.</p> <p>Many companies embark on convergence to enhance their security processes. That effort often involves finding ways to adopt traditional IT security measures to the device-oriented IoT environment that OT supports. There might also be security measures in place specifically designed to protect OT endpoint devices, so those methods and processes must be integrated with IT security. IoT security can be particularly tricky given the number and types of devices that are connected to the network -- increasing the potential attack surface.</p> <figure class="main-article-image full-col" data-img-fullsize="https://www.techtarget.com/rms/onlineimages/iti-integrating_it_and_ot-f.png"> <img data-src="https://www.techtarget.com/rms/onlineimages/iti-integrating_it_and_ot-f_mobile.png" class="lazy" data-srcset="https://www.techtarget.com/rms/onlineimages/iti-integrating_it_and_ot-f_mobile.png 960w,https://www.techtarget.com/rms/onlineimages/iti-integrating_it_and_ot-f.png 1280w" alt="Venn diagram showing the integration of IT and OT." height="387" width="559"> <figcaption> <i class="icon pictures" data-icon="z"></i>IT/OT convergence involves integrating two distinct networks, sets of data and teams. </figcaption> <div class="main-article-image-enlarge"> <i class="icon" data-icon="w"></i> </div> </figure> <p>Convergence can be introduced on different levels. It might be principally based on the following:</p> <ul class="default-list"> <li>Physical (hardware and environment) convergence of systems.</li> <li>Software convergence.</li> <li>Data convergence.</li> </ul> <p>In all cases, there will likely be a need for some degree of translation/integration, and in most cases, IT/OT convergence will require some degree of integration at all three levels.</p> </section> <section class="section main-article-chapter" data-menu-title="Benefits of IT/OT convergence"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Benefits of IT/OT convergence</h2> <p>The chief benefit of convergence is cost. Maintaining two separate networks is an expensive proposition. By merging networks, it's possible to reduce the amount of required networking gear, as some parts of the converged network will end up doing double duty, serving both the industrial and front-office sides of a business.</p> <p>A converged physical network also makes data sharing much easier, which, as noted, can benefit the processes running on both sides of the business. But it can also mean that data can be acted on more immediately, and <a href="https://www.techtarget.com/searchstorage/feature/5-tips-for-improving-data-storage-efficiency">data storage resources can be combined</a> for an additional economic benefit. IT and the business units it supports will be smarter when OT's real-time data is incorporated into their data sets for enhanced analysis, enabling practices such as just-in-time manufacturing and smarter supply chain management.</p> <p>The OT staff can combine sales and marketing data from IT with the voluminous data it collects to control manufacturing processes more efficiently. That way, production of best-selling products can be ramped up, while production of less popular products can be cut back.</p> <p>Although there are opportunities for cross-training to build a converged staff, organizations might still want some staff to specialize in OT or IT issues and devices, largely because so many of the devices that populate the OT world are unfamiliar to computer experts. The operational and health information these devices provide might be unique or industry-specific.</p> <p>With AI and machine learning becoming more prominent in applications across both IT and OT environments, it makes sense to integrate their capabilities to interpret and act on data more effectively. For example, TinyML is bringing machine learning capabilities to more IoT endpoint devices; integrating these capabilities with other AI-powered applications promises benefits for both IT and OT environments.</p> <p>AI applications also require that data achieves some level of homogeneity, so that the data -- regardless of where and how it was captured -- can potentially play a critical role in an AI system. That means industrial data that was traditionally rarely used outside of machinery control and monitoring environments might now have greater value, especially when combined with data from other sources, such as traditional IT systems (e.g., <a href="https://www.techtarget.com/searchcustomerexperience/definition/CRM-customer-relationship-management">CRM</a>, ERP, transactional apps). Centralizing data management is an imperative, as it is key to effective data governance and compliance.</p> <p>Connections to outside organizations have become important to both environments as well, so combining their networks and network connections can benefit them equally. For some IoT environments, such as energy distribution from power plants, outside connectivity is necessary. And traditional IT networks have long relied on remote connectivity -- often using cloud-based services -- to stay in touch with suppliers and customers.</p> <p>From an IT perspective, convergence means drastically extending the reach of installed systems and computing resources, which could create management issues, but might also present new opportunities for business expansion.</p> </section> <section class="section main-article-chapter" data-menu-title="Challenges of IT/OT convergence"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Challenges of IT/OT convergence</h2> <p>The challenges of implementing a converged IT/OT environment might be significant, as it will likely require changes to procedures in both technical disciplines. Some of the obstacles that organizations must address include the following:</p> <ul class="default-list"> <li><b>Scale of connectivity.</b> OT typically connects far more devices than an IT network. The sheer number of devices involved in a convergence effort must be considered before networks converge.</li> <li><b>Device inventory.</b> Everything that's connected to IT and OT networks must be accounted for to ensure that devices aren't left unsecured&nbsp;or orphaned. Communication among diverse devices must be tested and confirmed. This also suggests that IT and OT personnel develop a basic understanding of each other's equipment to enhance troubleshooting and remediation efforts.</li> <li><b>Firmware and other updates.</b> As part of its security efforts, IT tends to emphasize firmware and system software updates to help ensure that no vulnerabilities are exposed. OT, on the other hand, might host devices that have been used for many years and might have out-of-date software -- or might no longer be supported. Those issues must be resolved to ensure that endpoint devices don't present additional vulnerabilities.</li> <li><b>Encryption.</b> The converged systems should <a href="https://www.techtarget.com/whatis/feature/Mobile-device-encryption-How-it-works-and-how-to-enable-it">encrypt all communication between devices</a> and other processing resources, whether they are internal to the organization or provided by external services.</li> <li><b>Adjust networking systems for different types of data and transmission rates.</b> Systems are often tuned to the types of data they collect and transmit. For example, OT data is typically very small and voluminous, while IT data can range from small files to huge media files. Some tuning of network devices might be required, as data is bound to travel more and mix more frequently across a converged OT/IT environment.</li> <li><b>Hybrid storage systems.</b> In most cases, the data collected in IT and OT environments will be merged at some point. Combining edge storage with more traditional centralized storage can pose problems due to the differing data types that each environment supports. Differences in network bandwidth, data types, capacities, access frequency and data ingestion rates must all be resolved. IT groups with limited experience with cloud storage will face a steep learning curve, as IoT endpoint storage often relies on proximity to a cloud storage service.</li> <li><b>Merging data.</b> Gathering data from multiple sources is just the start of the data sharing process. The data is likely to be in different formats and adhere to different protocols, so some form of data management middleware will likely be needed to handle the disparities in the collected data.</li> <li><b>Ask for a software bill of materials. </b>An <a href="https://www.techtarget.com/whatis/definition/software-bill-of-materials-SBOM">SBOM</a> is a list of all the software components and dependencies that go into a device deployed in an IoT environment. Providing SBOMs is a newer practice by product vendors, but it's a key step in addressing issues related to the variety and age of IoT devices.</li> </ul> <p>Security might be at the top of the IT/OT convergence list of challenges, but ironically, it can also loom as a significant potential benefit because consolidated network management is likely to make securing network resources easier and more effective.</p> <p>For IT, security activities typically focus on the network infrastructure that transports data and the storage and memory systems where applications and data reside, either permanently or transiently. OT must also secure a similar set of components, but the task is compounded by the sheer number of endpoint devices that can connect to operational networks. Those endpoints can be a mix of legacy and new devices with varying needs for patches and updates. So, job one might be using management software to establish an accurate inventory of devices and to methodically update it as needed.</p> <p>In a converged world, security should be bolstered by modern firewall technologies, such as <a href="https://www.techtarget.com/searchsecurity/definition/Web-application-firewall-WAF">web application firewalls</a> and VPNs, which encrypt data in motion across remote network connections and add a layer of protection against a variety of cyberattacks.</p> <p>Backing up data is a crucial part of security and of maintaining business processes to avoid disruptions for both IT and OT, but the location and nature of the data to be protected differ. Backup and disaster recovery applications and processes must work effectively in both environments and efficiently access and <a target="_blank" href="https://wesolve.tech/blog/endpoint-backup-protecting-data-on-every-device/" rel="noopener">back up</a> a range of endpoint storage devices.</p> <p>There are, of course, also human elements that must be considered for a convergence effort to succeed. The effects of organizational convergence can be just as important as systems convergence:</p> <ul class="default-list"> <li>Traditionally disparate teams will be involved, but despite the differences, there will be significant overlapping responsibilities that must be sorted out for convergence.</li> <li>All parties, from the C-suite to the data center floor, will have to come to an understanding of common goals that might be beyond the scope of what they had experienced when IT and OT were separate entities.</li> <li>The inevitable redistribution of responsibilities will likely change reporting structures as well as personal goals and expectations.</li> </ul> <p><i>Rich Castagna has been a high-tech journalist for more than 30 years. Rich worked at TechTarget for 15 years, overseeing technical coverage and content creation as vice president of editorial. Previous roles include executive editor of ZDNet Tech Update and CNET Enterprise, and editor-in-chief of Windows Systems magazine. </i></p> </section> IT and OT have long been independent disciplines, but technical developments and new processing requirements -- particularly for AI and IoT -- are compelling IT/OT convergence. https://cdn.ttgtmedia.com/rms/onlineimages/machine_learning_g1307219089.jpg https://www.techtarget.com/it-infrastructure/tip/IT-vs-OT-Key-differences-explained Thu, 20 Aug 2026 09:15:00 GMT <p>A major GitHub outage this week left some users unable to deploy code with GitHub Actions for hours, but switching to an alternative would be an immense undertaking, especially for companies with a large code footprint.</p> <p>GitHub's incident page first reported an investigation just after 9:30 a.m. EST Aug. 17, then degraded performance for services including API requests, GitHub Actions and webhooks. Just after 10 a.m., the incident page reported high error rates -- around 20% -- for web experiences and API traffic, along with a 50% error rate for archive and raw repository content downloads. These high error rates then spread to SAML and OIDC authentication services. About an hour after the incident began, GitHub Copilot, pull requests, GitHub Issues and GitHub Actions were experiencing degraded availability.</p> <p>These services include critical software deployment capabilities, including project management and CI/CD system triggers. <a target="_blank" href="https://www.techtarget.com/it-infrastructure/news/366623845/New-GitHub-Copilot-agent-edges-into-DevOps" rel="noopener">GitHub Copilot</a> has become a central component of agentic DevSecOps functions on the platform.</p> <p>An incident post-mortem added at 5:15 p.m. EST Aug. 17 <a href="https://www.githubstatus.com/incidents/zkxwbgr0cnmx">said the outage</a> began with "network saturation on load balancers in Central US due to a new peak in traffic." All told, GitHub Copilot suffered a six-hour-and-44-minute outage, while GitHub API requests, Issues, pull requests and Actions were down for more than two hours, according to a GitHub <a href="https://www.githubstatus.com/">status page</a>.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> That outage had me fuming and considering an alternative. All of our work ground to a halt </figure> <figcaption> <strong>Steve Koelpin</strong>Principal AI observability engineer, Fortune 50 company </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>It's not the first time that a sharp uptick in AI-agent-generated traffic has <a href="https://www.techtarget.com/it-infrastructure/news/366644857/GitHub-Copilot-desktop-released-amid-reliability-and-pricing-concerns">caused reliability problems for GitHub</a> this year. GitHub blog posts <a href="https://github.blog/news-insights/company-news/an-update-on-github-availability/">in April</a> and again <a href="https://github.blog/news-insights/company-news/github-availability-report-july-2026/">Aug. 12</a> called the severity, frequency and duration of incidents unacceptable and pledged improvements.</p> <p>But the scope of the Aug.17 outage was broader for some users than previous incidents had been.</p> <p>"That outage had me fuming and considering an alternative," said Steve Koelpin, principal AI observability engineer at a Fortune 50 company, of the Aug. 17 incident. "All of our work ground to a halt. We couldn't deploy, and our Actions were down."</p> <p>Previous reliability incidents hadn't had such a broad impact, Koelpin said.</p> <p>"For prior outages, service degraded, but the rest of the loop still worked, so you could route around it," he said. "Monday hit everything at once: pull requests, Actions, webhooks, SAML/OIDC, for roughly seven and a half hours. Git operations stayed up, so we could push [code], but couldn't review, merge or deploy. That's an inconvenience becoming a full stop."</p> <p>Multiple customers other than Koelpin told TechTarget that errors and timeouts brought deployments to a full halt.</p> <p>"We've been severely impacted. [It was] impossible to open or merge a PR for several hours yesterday," said Pascal Gandilhon, head of observability for Nexthink, Inc., makers of digital employee experience management software in Lausanne, Switzerland.</p> <section class="section main-article-chapter" data-menu-title="Costs of leaving GitHub still outweigh outage pain"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Costs of leaving GitHub still outweigh outage pain</h2> <p>Gandilhon said there hasn't been any talk of switching to a different platform at his company.</p> <p>"We are deeply embedded with GitHub, but if this happens too often, for sure developers will start to complain," he said.</p> <p>Koelpin said he began to investigate alternatives this week, but was stymied by the amount of time and effort it would take to move.</p> <p>"The lock-in is Actions," he said. "Every workflow pulls Marketplace Actions that only exist on GitHub, and there's no equivalent catalog elsewhere. You rewrite all of it by hand, rebuild the runners, then redo branch protections, CODEOWNERS [files], SAML/SCIM, and every webhook feeding downstream. This is months of work, and none of it adds value that we don't have today."</p> <p>Another user affected by the actions agreed that Actions makes it difficult to migrate away from GitHub.</p> <p>"Migrating anywhere else means I have to figure out what to do with my build pipelines, and GitHub Actions has network effects," said Ricardo Torres, senior staff platform engineer at digital resiliency company Veeam Software. "Buildkite has a GitHub Actions runner, but again, I've got to still pay other migration costs. I've yet to see any of the new contenders really take on GitHub Actions."</p> </section> <section class="section main-article-chapter" data-menu-title="Clock ticking for GitHub as new rivals emerge"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Clock ticking for GitHub as new rivals emerge</h2> <blockquote class="twitter-tweet"> <p lang="en" dir="ltr">We were going to ship this earlier, but GitHub was down <a href="https://t.co/uxsBqt3yKl">https://t.co/uxsBqt3yKl</a></p> — matt palmer (@mattyp) <a href="https://x.com/mattyp/status/2089400829933666599?ref_src=twsrc%5Etfw">August 17, 2026</a> </blockquote> <p> <script src="https://platform.x.com/widgets.js"></script> </p> <p>Even if they don't migrate codebases and pipelines to a different platform, some customers affected by GitHub outages might consider setting up a secondary platform on Atlassian's Bitbucket or GitLab to continue operations in the event of another incident. But these alternatives don't amount to a full replacement for GitHub for now, according to Koelpin and Torres.</p> <p>"Bitbucket isn't parity. You'd trade capability for uptime. GitLab is the only real alternative," but it can't match GitHub Actions Marketplace integrations, Koelpin said.</p> <p>Meanwhile, newer competitors are clearly attempting to seize the moment of vulnerability for GitHub with new agentic DevSecOps services, including Cursor's Origin code hosting service, which launched Aug. 17. Cursor's agentic IDE saw a meteoric rise in popularity over the last two years, culminating in usage by 64% of Fortune 500 companies and more than 50,000 enterprises, according to the company's website. Cursor was bought by SpaceX for $60 billion in a deal that closed this month.</p> <p>Cursor Origin remains in early beta, but it's rolling out to all paid plans, and "agent-native features ship soon," according to a Cursor <a target="_blank" href="https://cursor.com/changelog/origin-code-hosting" rel="noopener">blog post</a>.</p> <p>For now, the market for agentic DevSecOps remains GitHub's to lose, said Jason Andersen, an analyst at Moor Insights &amp; Strategy. But its lead won't last indefinitely.</p> <p>"The issue is that while there are some alternatives like GitLab and even some new lifecycle management offerings from cloud providers, GitHub's lead is huge and these options are catching up," Andersen said.&nbsp;"My guess is that, over time, other agentic offerings will continue to get better and GitHub will need to compete more vigorously."</p> <p>A GitHub spokesperson did not comment beyond sending a link to the incident post-mortem update.</p> <p><em>Beth Pariseau, senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism. Have a tip?&nbsp;</em><a target="_blank" href="mailto:bpariseau@techtarget.com?subject=News%20tip" rel="noopener"><em>Email her</em></a><em> or connect on </em><a target="_blank" href="https://www.linkedin.com/in/bethpariseau" rel="noopener"><em>LinkedIn</em></a><em>.</em></p> </section> Degraded availability on GitHub's network prevented some customers from using Copilot and deploying with GitHub Actions. Frustrated users found there's no easy alternative. https://cdn.ttgtmedia.com/rms/onlineimages/disaster_recovery_a78784722.jpg https://www.techtarget.com/it-infrastructure/news/366649459/GitHub-outage-had-users-weighing-options-but-finding-few Wed, 19 Aug 2026 13:38:00 GMT <p>AI depends on more than powerful models and accessible compute resources. As enterprises integrate AI into products, services and operations, decisions about storage, compute, governance and data management will ultimately determine the success of scaling AI initiatives.</p> <p>The successful integration of AI requires organizations to balance interconnected infrastructure considerations with the alignment of technical capabilities to business objectives. Although challenges are inevitable, strategic infrastructure planning can improve performance, control costs and create sustainable competitive advantages. To harness these emerging capabilities, <a href="https://www.techtarget.com/enterprise-software/feature/Routine-IT-updates-now-shape-enterprise-governance">enterprise governance</a> and operational efficiency become the rate-limiting factors that balance opportunity and costs.</p> <section class="section main-article-chapter" data-menu-title="4 infrastructure considerations for AI integration"> <h2 class="section-title"><i class="icon" data-icon="1"></i>4 infrastructure considerations for AI integration</h2> <p>Successfully integrating AI requires more than deploying powerful models. Organizations must consider four interconnected infrastructure priorities: balancing storage and compute requirements; establishing governance frameworks for machine learning operations (MLOps); engineering data pipelines that support evolving AI processing needs; and modernizing legacy systems through incremental change.</p> <h3>Balancing storage, compute and performance requirements</h3> <p>Balancing scalable platforms that decouple storage from compute is vital. Modern AI workloads introduce performance variability across hardware and require a mix of high-speed (hot) and archival (cold) storage, as well as cluster-based or service-based compute architectures. Elastic demands and uneven wear of on-demand compute resources for <a href="https://www.techtarget.com/it-infrastructure/answer/AIOps-vs-MLOps-How-are-they-different">AI and MLOps</a> are primary factors for IT to consider. Balanced scaling of infrastructure storage and compute clusters optimizes resource use in elastic and edge computing contexts.</p> <p>Throughput, latency, scalability and resiliency are key metrics for measuring storage performance. Scaling storage to meet the demand for AI workloads without contributing to technical debt requires a balanced approach to infrastructure transformations. Scaling compute, such as GPUs or tensor processing units, without balancing I/O and storage can lead to bottlenecks in hardware utilization. Upfront investment in a high-throughput system interconnect helps future-proof infrastructure.</p> <p>New server chips with increased throughput from <a target="_blank" href="https://ieeexplore.ieee.org/document/11429758" rel="noopener">MRDIMM multi-rank memory</a> technology, such as the Intel 18A series, must balance the performance increase against system throughput to avoid bottlenecks. Additionally, while high-performance clusters can use such chips, AI inference and edge computing workloads will not necessarily warrant such a high-power investment and are poised to be the greatest growth area, according to <a target="_blank" href="https://download.intel.com/newsroom/2026/Data-Center/Computex-2026-Data-Center-Group-Press-Deck.pdf" rel="noopener">Intel Data Center Strategy projections</a> at Computex in 2026.</p> <p>Careful forecasting of requirements for future business processes and architecture needs is a tripartite problem involving storage, compute and <a target="_blank" href="https://www.techtarget.com/it-infrastructure/tip/Navigating-energy-management-strategies-in-AI-data-centers" rel="noopener">energy demands</a>. These classic systems issues are complicated by site-based edge and inference requirements for real-time products across service areas, depending on data center-edge round-trip time and worst-case processing delays.</p> <h3>Building governance frameworks for AI workloads</h3> <p>Data governance in AI now extends far beyond traditional access control. ML workflows contain additional governance tasks such as lineage tracking, role-based permissions for model modification and policy enforcement over how data is labeled, versioned and reused. This includes data set documentation, drift tracking and <a href="https://www.techtarget.com/whatis/feature/12-of-the-best-large-language-models">large language model-specific controls</a> over prompt inputs and generated outputs.</p> <p>Governance frameworks that support continuous learning cycles are more valuable: Every inference and user correction can become training data. An inference run twice is an inference twice paid for. Systems that log, audit and review how these feedback loops affect downstream behavior stand to benefit the most. Without structured oversight, model outputs risk reinforcing bias or violating compliance norms. Through metadata schemas, compliance monitors and ML-aware policy engines, IT infrastructure becomes an opportunity to embed governance at the workflow layer.</p> <p>In managed and AI service strategies, governance is more pivotal as multivendor configurations increase complexity. Governance and workflow efficiency are management's tools to throttle insight costs in provisioned AI use. As each API call becomes the new variable cost of labor associated with business process fulfillment, <a target="_blank" href="https://www.techtarget.com/it-infrastructure/feature/Cloud-cost-management-strategies-for-upcoming-volatile-decade" rel="noopener">cost optimization strategies</a> for AI workloads are critical to maintaining business strategy.</p> <h3>Engineering data pipelines for modern AI systems</h3> <p>AI pipelines demand a seamless workflow of data ingestion, transformation, model inference and feedback loops. As models become more stateful and retain context over time, pipelines must support real-time, memory-intensive operations while reducing cold starts and expensive idle time. AI workflows are moving toward hybrid stateful models that can handle ongoing, contextual tasks and smaller stateless, single-pass processing. Holding one or more training models in memory to perform ingestion, processing and output tasks enables monitoring and continuous model training and deployment, while inference might only require data to be output to the edge.</p> <p>Designing governance and management practices around data pipelining enables elastic resource allocation for AI processing and matches workflow streams to business processes. This approach can create uneven resource demand and component turnover for IT teams to monitor or cause management pressure to shift toward an <a href="https://www.channeldive.com/news/archive-Guide-to-building-and-executing-an-MSP-business-model/815806/">MSP model</a> of resource provision.</p> <p>Stateful model tasks are highly storage- and compute-intensive and create data-intensive workflows. Salesforce <a target="_blank" href="https://ieeexplore.ieee.org/document/11331297" rel="noopener">has developed</a> an inference architecture focused on platform-agnostic containerization strategies to address the unique AI production issues of throughput bottlenecks, expensive idle processor time and cold-start latency; using these requires governance and well-organized workflows.</p> <h3>Modernizing legacy infrastructure for AI integration</h3> <p>A successful legacy transition focuses on modular upgrades and incremental improvements that reduce risk while enabling AI-driven capabilities to operate alongside existing legacy systems. Containerizing model training and inference environments enables parallel operation and rollback isolation, mitigating risk while sandboxing deployments for departments to train on and adapt to new workflows.</p> <p>Although each transformation creates unique technical challenges, they are deeply interconnected. Decisions about storage architectures <a href="https://www.techtarget.com/searchenterpriseai/tip/Tools-and-techniques-for-optimizing-AI-data-pipelines">affect data pipelines</a>. Governance requirements influence infrastructure design. Legacy modernization strategies determine how quickly organizations can adopt emerging AI capabilities. Engineers must balance these considerations simultaneously rather than treat them as independent initiatives.</p> </section> <section class="section main-article-chapter" data-menu-title="How to manage infrastructure evolution strategically"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How to manage infrastructure evolution strategically</h2> <p>Transforming IT infrastructures for AI workflows and pipelines can be a high-risk undertaking. Gradual refactoring enables targeted infrastructure changes that deliver meaningful, high-impact performance improvements for select business processes. Managing for future-proofed architecture and elastic scalability while maintaining a targeted focus on resource-intensive processes that deliver the greatest returns can be challenging to define. The process will contribute to more transparent data governance and a deeper understanding of business processes.</p> <p><a name="_Hlk147342280"></a>The adoption of generative AI and hybrid AI continues to expand among companies with over <a target="_blank" href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" rel="noopener">$500 million in revenue</a>. Data infrastructures and C-suite strategies remain primary hurdles and accelerators for AI adoption. Mobilizing for shifts in operations to continuous integration and continuous training requires Elasticsearch and high-capacity storage to realize the strategic advantage that AI technologies promise. Robust data governance programs and worker upskilling are critical management initiatives, but without a data infrastructure that supports the ongoing use of AI technologies in business functions, the best strategy will fail to scale.</p> </section> <section class="section main-article-chapter" data-menu-title="Infrastructure decisions will determine AI outcomes"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Infrastructure decisions will determine AI outcomes</h2> <p>Organizations that successfully integrate AI infrastructure require executive sponsorship, established governance and infrastructure investments that consider targeted outcomes. Continuous improvement, <a href="https://www.techtarget.com/it-infrastructure/feature/How-AI-in-training-and-development-can-bridge-IT-talent-gaps">development and training</a> offer operational approaches amid mounting regulatory and market pressures to adopt AI features in product delivery. As AI workflows evolve, enterprises must align infrastructure considerations with technical modernization, model governance, risk management and workforce readiness. More than a compute problem, AI integration is an architecture, governance and resiliency challenge, and successful integration depends on the underlying infrastructure.</p> <p><i>Hardik Chawla is a senior product manager with nearly eight years of experience in digital product management. He is responsible for supply chain optimization and technology, driving development of B2B platforms, API-first architecture and AI/ML-driven products. He holds a bachelor's degree in electrical engineering and an MBA in technology management and strategy from the UCLA Anderson School of Management. Connect with Hardik on </i><a target="_blank" href="https://www.linkedin.com/in/hchawla94/" rel="noopener"><i>LinkedIn</i></a><i>.</i></p> </section> AI adoption can expose weaknesses across the enterprise. A strategic approach to infrastructure modernization can help organizations manage risk, efficiency and long-term growth. https://cdn.ttgtmedia.com/rms/onlineimages/ai_g1183318665.jpg https://www.techtarget.com/it-infrastructure/post/Balancing-infrastructure-priorities-for-AI-integration Wed, 19 Aug 2026 09:18:00 GMT <p>As of Aug. 2, all text made with Claude will bear a&nbsp;word-pattern-based watermark that is detectable by machines but invisible to human eyes.&nbsp;</p> <p>Does it matter if the image you see or text you read was created by a human or generative AI? In some contexts, it matters a great deal. That is why the EU introduced<a href="https://artificialintelligenceact.eu/article/50/" target="_blank" rel="noopener"> Article 50: Transparency Obligations for Providers and Deployers of Certain AI Systems</a> on Aug. 2, prompting <a href="https://www.techtarget.com/it-strategy/feature/Weekly-news-roundup-Claude-watermark-controversy-and-Nvidia-500-billion-deal">Anthropic to announce</a> the watermarking strategy.</p> <p>Anthropic has committed to applying its watermark to output from supported models wherever Claude is offered and to add marking support to Claude models released before Anthropic introduced the <a href="https://www.techtarget.com/ai/definition/AI-watermarking">watermark</a>.</p> <section class="section main-article-chapter" data-menu-title="How does Claude’s text watermarking work?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How does Claude’s text watermarking work?&nbsp;</h2> <p>Claude’s approach operates at the model level, meaning it will be present regardless of the specific product or surface overlaid on Claude.</p> <p>Claude creates the watermark by replacing the&nbsp;LLM’s random <a href="https://www.techtarget.com/it-infrastructure/news/366642661/Atlassian-MCP-updates-take-aim-at-AI-token-usage">token selection</a> with a specific, checkable pattern set by a secret key. The possessor of the key can determine if the text matches such a pattern to flag what shows strong indications of having been composed by Claude.&nbsp;</p> <p>When Claude generates text, it interlaces an imperceptible watermark directly into the text itself. According to Anthropic, the inclusion of the watermark does not change the meaning, readability or quality of Claude's response. It persists through copying and pasting, and some editing.</p> <h3>How Claude marks supported filetypes</h3> <p>Claude's approach to watermarking generated file types such as a .svg, .png, or .jpg differs from the text-based approach. &nbsp;</p> <p>In these cases, it attaches signed provenance metadata that follows the Coalition for Content Provenance and Authenticity (<a href="https://www.techtarget.com/whatis/definition/Coalition-for-Content-Provenance-and-Authenticity-C2PA">C2PA</a>) open standard to establish content provenance. A signed metadata label signals that Claude processed the file and provides a way to detect whether the file has been tampered with.</p> <p>The text watermark is not a piece of metadata or additional characters included in the text; it's a pattern in the text itself.</p> </section> <section class="section main-article-chapter" data-menu-title="How does it compare to other watermarking approaches?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How does it compare to other watermarking approaches?&nbsp;</h2> <p>Anthropic did not invent this approach to <a href="https://www.techtarget.com/it-infrastructure/tip/Tips-for-the-governance-of-AI-generated-and-synthetic-data">flagging AI-generated content</a>. Google DeepMind introduced <a href="https://deepmind.google/models/synthid/" target="_blank" rel="noopener">SynthID</a> -- its version of a proprietary statistical watermark that lives in generated text -- in 2023. Both are statistical/model-level approaches in which a hidden signal influences token selection to establish a pattern discernible to the detector in possession of the key. They also both persist through copy-paste.</p> <p>Claude's approach and SynthID also share the same weaknesses:&nbsp;</p> <ul class="default-list"> <li><b>Rely on longer texts to establish the pattern. </b>This approach should work on an article or blog post but likely would not flag a short social media post.</li> <li><b>Requires word selection.</b> Claude might not generate the pattern for text that doesn't involve any unique word selection because it is made up solely of code, numbers or even lists of dates or proper names.</li> <li><b>Significant editing can break them.</b> Even though the pattern is not explicitly stripped out by a copy and paste, if the text gets altered significantly in editing, the pattern might fade away.&nbsp;</li> </ul> <p>The key differences between SynthID and Claude's approach are a matter of model specificity and scope.</p> <ul class="default-list"> <li><b>Specificity.</b> Google's SynthID detector is based on its own secret key which only applies to Google-generated watermarks. It cannot detect Claude's. The converse is true as well. Anthropic's key only works for Claude, not for any other model.&nbsp;</li> <li><b>Scope. </b>SynthID works for a wide range of files, including images, audio and video. In contrast, for anything other than text, Anthropic has stated that it will rely on the <a href="https://c2pa.org/" target="_blank" rel="noopener">C2PA</a> open standard instead of the token sampling approach. There is no proprietary <a href="https://www.techtarget.com/cybersecurity/answer/Symmetric-vs-asymmetric-encryption-Understand-key-differences">secret key</a> involved in the C2PA approach. Instead, it attaches identifying metadata to the file.</li> </ul> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">How C2PA compares</h3> <p>Anyone can strip C2PA metadata out with minimal effort.&nbsp;But it offers two significant upsides over the token sampling approach:</p> <ul type="disc" class="default-list"> <li>The open standard enables independent verification without access to a secret key.</li> <li>The metadata can survive extensive revision. &nbsp;&nbsp;&nbsp;&nbsp;</li> </ul> </div> </div> <ul class="default-list"></ul> </section> <section class="section main-article-chapter" data-menu-title="Yay or nay on Anthropic’s announcement?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Yay or nay on Anthropic’s announcement?&nbsp;</h2> <p>While transparency is generally considered a good thing, the reaction to Claude's announcement was not universally positive. Anthropic has taken note of some responses and addressed them in speaking with journalists and directing people to its FAQ in an X <a href="https://x.com/AnthropicAI/status/2088343978873966687" target="_blank" rel="noopener">post</a> on Aug. 14.&nbsp;</p> <p>A good portion of those 1,900 comments represent people firmly in the nay group. Even some who say AI detection is both not new and necessary are not attributing the best intentions to Anthropic's move.&nbsp;</p> <p>Donn Felker, a software developer, pointed out on <a target="_blank" href="https://x.com/donnfelker/status/2087180167907054066" rel="noopener">X</a> that the watermark could help AI models recognize and avoid training on AI-generated content – a circumstance that can degrade and homogenize outputs.<i> </i></p> <p>Despite espousing the potential benefits of the approach, a follow-up response to a comment on his post comes off as more critical:&nbsp;</p> <p>"Yeah the Anthropic anti-bad guy stance isn't something I really believe in anymore. Especially after dealing with their enterprise sale process. They're very good at getting every penny they can."&nbsp;</p> </section> <section class="section main-article-chapter" data-menu-title="The connotations of a secret key"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The connotations of a secret key&nbsp;</h2> <p>On Aug. 11, <a href="https://x.com/bgurley/status/2087335941216272548" target="_blank" rel="noopener">Bill Gurley wrote on X</a>: "The word "watermark" comes from photos and is visible by all. This is only identifiable by Anthropic. Once again, they are judge, jury and prosecutor. Lots of reasons to not use their products."</p> <p>Anthropic must have recognized its exclusive possession of the key as a valid concern. A representative told Business Insider <a target="_blank" href="https://www.businessinsider.com/anthropic-claude-text-watermark-concerns-tech-community-answers-2026-8" rel="noopener">that</a> it intended to release an API "that will allow users and third parties to check text for Claude's watermark themselves." That would appear to be a step in the right direction for any business that claims to be acting in the name of transparency.</p> </section> <section class="section main-article-chapter" data-menu-title="A lack of nuance in Claude's watermarking"> <h2 class="section-title"><i class="icon" data-icon="1"></i>A lack of nuance in Claude's watermarking&nbsp;</h2> <p>Critics of AI detectors often point out that they can flag wholly human-created text as containing signals of AI-generation due to patterns that LLMs learned to emulate. As Andrea Saez explains <a href="https://dreasays.substack.com/p/putting-claudes-watermarking-to-the" target="_blank" rel="noopener">in</a> the blog "Putting Claude's watermarking to the test," one should not mistake signals for absolute proof of AI authorship. Due to the common use of particular phrases and repeated words, it is possible that a detector might pick up false positives, even from a text that is 100% human-generated.&nbsp;</p> <p>Real life examples of this abound.</p> <p>Theo Browne, a software engineer and entrepreneur who <a target="_blank" href="https://www.youtube.com/watch?v=Be-NqsW-wuk" rel="noopener">posts</a> popular tech content under the handle t3.gg, mentioned that contemporary <a href="https://medium.com/ai-ai-oh/i-ran-shakespeare-through-an-ai-checker-74ce20edc62f" target="_blank" rel="noopener">content creators</a> often have their content flagged as AI&nbsp;because the LLMs were trained on their content.</p> <p>The fact that Claude seems to apply the watermark to everything it touches exacerbates the problem. Someone who composes a text and merely uses Claude to proofread or translate it would receive the same label as someone who had Claude create the text from scratch.&nbsp;</p> <p>Simon<u> </u>Smith, executive vice president of generative AI at Klick Health, addressed that on X on <a href="https://x.com/_simonsmith/status/2088987931578949824" target="_blank" rel="noopener">Aug. 16</a>: "To clarify, I don’t think most people are upset by HOW Anthropic is watermarking. It’s that they’re stamping "Generated by Claude" on everything, not just fully AI-generated articles, but literally everything Claude helps with, which goes far beyond the EU AI Act’s requirements."</p> <figure class="main-article-image full-col" data-img-fullsize="https://www.techtarget.com/rms/onlineimages/smith-h.jpg"> <img data-src="https://www.techtarget.com/rms/onlineimages/smith-h_mobile.jpg" class="lazy" data-srcset="https://www.techtarget.com/rms/onlineimages/smith-h_mobile.jpg 960w,https://www.techtarget.com/rms/onlineimages/smith-h.jpg 1280w" alt="Smith on a lack of nuance in Claude watermark" data-credit="Ariella Brown"> <div class="main-article-image-enlarge"> <i class="icon" data-icon="w"></i> </div> </figure> <p>Smith's concern was echoed in the subhead of an Ars Technica <a href="https://arstechnica.com/tech-policy/2026/08/claudes-new-scarlet-letter-watermark-is-invisible-for-now/" target="_blank" rel="noopener">article</a> on the news, "Claude’s new Scarlet Letter watermark is invisible -- for now." The subhead notes: "The mark flags anything Claude processed, even human writing it only edited."</p> <p>In the Business Insider article that gave Anthropic the opportunity to respond to concerns, the unnamed company representative dismissed Smith's concern about light Claude use triggering the statement. "A watermark shows that Claude processed text, not necessarily that it wrote it, and that the mark can remain after Claude has proofread, translated, or summarized content."</p> <p>One wonders if the Anthropic representative is being deliberately obtuse here. Smith's and Ars Technica's argument is that using one label for any text touched by Claude is misleading.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Claude goes way beyond the letter of the law</h3> <p>Even though Anthropic claims that its watermarking is necessary to remain in compliance with the <a href="https://www.techtarget.com/searchenterpriseai/news/366646620/EU-AI-Act-compliance-deadline-is-here-What-to-watch">EU AI Act</a>, it goes much further than required. Content that the EU law allows to remain label-free would still get the mark.</p> <p>Even wholly AI-generated text can be exempt from the labeling requirement, as is the case for works of fiction or even marketing copy.&nbsp;</p> <p>Only text intended "to inform the public on matters of public interest" is required to show AI labels (and there's even a loophole for that if a human editor signs off on it).</p> </div> </div> <p></p> </section> <section class="section main-article-chapter" data-menu-title="Concerns about copyright, data privacy and quality questions"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Concerns about copyright, data privacy and quality questions</h2> <p>There are additional copyright concerns as well. For example, the UK-based software engineering coach John Crickett posted <a target="_blank" href="https://x.com/johncrickett/status/2087090150299881645" rel="noopener">on</a> X:&nbsp;</p> <p style="padding-left: 40px;"><i>Code from Claude Code is now watermarked. I wonder where that leaves the copyright ownership? If the code [is] AI generated, copyright cannot be claimed. If it's watermarked as AI generated, how can the author claim it had enough human input to claim copyright? How could this affect the valuation of tech companies?</i></p> <p>Steven Sinofsky wrote on X that he was <a target="_blank" href="https://x.com/stevesi/status/2087219853740720385" rel="noopener">worried</a> about digital privacy: "The real issue is data retention and your right to private thoughts free of a digital trail."</p> </section> <section class="section main-article-chapter" data-menu-title="Is more authentication better?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Is more authentication better?&nbsp;</h2> <p>The question that arises from this is: Does erring on the side of more watermarking make it impossible to pass off AI-generated content as human-generated? The answer appears to be no.</p> <p>Saez, Ars Technica and Browne all point out that it is possible to obliterate Claude's watermark by pasting the text into a rival generative AI and telling it to edit it. "I just cannot fathom any method where you can watermark text that isn't either incredibly expensive to detect or incredibly cheap to work around," Browne said in a YouTube video.</p> <p>Accordingly, he predicts "that we'll get over this watermarking phase relatively quickly," Instead he suggests focusing on marking authenticated human content. &nbsp;</p> <p><i>Ariella Brown is a technology journalist with experience covering AI, blockchain, IoT and cybersecurity.</i></p> </section> Anthropic now watermarks text generated or processed by Claude. Here's how the hidden pattern works, its limitations and why the approach has drawn criticism. https://cdn.ttgtmedia.com/visuals/digdeeper/1.jpg https://www.techtarget.com/it-infrastructure/opinion/Does-Claudes-watermark-solve-the-AI-transparency-problem Tue, 18 Aug 2026 17:49:00 GMT <p>Many organizations remain unprepared for the rapid growth of AI-generated content and synthetic datasets across enterprise environments, leaving them equally unprepared to govern that data effectively and within compliance boundaries.</p> <p>Governance frameworks are lagging behind AI, even as AI-generated and other algorithmically&nbsp;generated <a href="https://www.techtarget.com/it-strategy/definition/What-is-synthetic-data-Examples-use-cases-and-benefits">synthetic data</a> permeate across business functions. Governance is now a strategic business issue, not just an IT concern. Executives should establish governance best practices before operational and regulatory complexity increases. IT leaders investing in AI need to construct an AI data lifecycle policy and establish scalable governance.</p> <section class="section main-article-chapter" data-menu-title="Why AI-generated and synthetic data require governance"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Why AI-generated and synthetic data require governance</h2> <p>AI-generated data and synthetic data differ fundamentally from traditional data. Traditional data originates from real-world business activities, customer interactions, transactions, sensors or human-created content. It is the data generated by years of doing business.</p> <p>AI-generated data is produced by machine learning models, including text, images, code, audio or analytics. It is often derived from prompts or existing data. For enterprises, AI-generated data accelerates content creation, software development, customer support and decision-making.</p> <p>Synthetic data is artificially created to statistically resemble real data without directly representing actual individuals or events. Synthetic data enables a host of practices, such as AI model training, <a href="https://www.techtarget.com/searchbusinessanalytics/tip/Synthetic-data-vs-real-data-for-predictive-analytics">software testing</a>, analytics and data sharing while reducing privacy risks, protecting sensitive information and improving compliance. Together, they expand opportunities for innovation but require stronger governance over provenance, quality, security and lifecycle management.</p> <figure class="main-article-image full-col" data-img-fullsize="https://www.techtarget.com/rms/onlineimages/when_to_use_synthetic_data-f.png"> <img data-src="https://www.techtarget.com/rms/onlineimages/when_to_use_synthetic_data-f_mobile.png" class="lazy" data-srcset="https://www.techtarget.com/rms/onlineimages/when_to_use_synthetic_data-f_mobile.png 960w,https://www.techtarget.com/rms/onlineimages/when_to_use_synthetic_data-f.png 1280w" alt="When to use synthetic data. " data-credit="TT" height="347" width="560"> <figcaption> <i class="icon pictures" data-icon="z"></i>Four synthetic data use cases are based on data accessibility and representativeness: edge cases, model validation, data scarcity, and privacy and security. </figcaption> <div class="main-article-image-enlarge"> <i class="icon" data-icon="w"></i> </div> </figure> <p>These data sources are increasingly important to business innovation, but they include significant risks for IT leaders to mitigate.</p> </section> <section class="section main-article-chapter" data-menu-title="How machine-generated content creates risk"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How machine-generated content creates risk</h2> <p>Unmanaged AI-generated content presents operational, legal and <a href="https://www.techtarget.com/searchSecurity/tip/Data-sovereignty-compliance-challenges-and-best-practices">compliance risks</a> that cannot be ignored. These risks span the organization, making them a strategic governance imperative.</p> <p>Specific risk examples include:</p> <ul class="default-list"> <li><b>Data provenance and authenticity</b>. Organizations must be able to demonstrate where AI-generated content originated, which model produced it, when it was created and whether it has been modified. Without provenance, trust and auditability suffer.</li> <li><b>Regulatory compliance</b>. AI-generated and synthetic data may still fall under regulations such as GDPR and the <a href="https://www.techtarget.com/ai/opinion/Everything-you-need-to-know-about-the-new-EU-AI-Act">EU AI Act</a> if personal information is involved or if AI outputs impact regulated decisions. Governance enables transparency, traceability and accountability.</li> <li><b>Retention and deletion obligations</b>. AI-generated content should follow the same retention, archival and <a href="https://www.techtarget.com/searchDataBackup/tip/Increase-backup-efficiency-with-a-data-destruction-policy">disposal policies</a> as other business records. Unmanaged data increases <a href="https://www.techtarget.com/searchdatacenter/tip/AI-capacity-planning-Balancing-flexibility-performance-and-risk">storage costs</a> and compliance risks.</li> <li><b>Intellectual property</b>. AI outputs may contain copyrighted or proprietary material, creating ownership disputes or licensing concerns.</li> <li><b>Third-party AI vendor risk</b>. External AI platforms can introduce contractual, <a href="https://www.techtarget.com/searchcloudcomputing/tip/A-data-sovereignty-primer-for-cloud-admins">data residency</a>, security and privacy risks. Organizations should define vendor governance requirements and understand how providers store, use and retain submitted data.</li> </ul> <p><a href="https://www.techtarget.com/searchsecurity/tip/Beyond-the-perimeter-The-shift-to-data-centric-protection">Enterprises face challenges</a> with proving origin and authenticity, yet they must be prepared to undergo audits and prove regulatory compliance. Building an AI lifecycle policy and governing data deliberately helps ensure these obligations are met.</p> </section> <section class="section main-article-chapter" data-menu-title="Building an AI data lifecycle policy"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Building an AI data lifecycle policy</h2> <p>A practical AI data lifecycle policy should establish governance from the moment AI-generated or synthetic data is created until it is securely disposed of. All aspects of the data's lifecycle require management to ensure privacy, compliance and auditability.</p> <p>Specific lifecycle events include:</p> <ul class="default-list"> <li><b>Inventory AI-generated data</b>. Identify where AI-generated and synthetic data is created, processed, shared and stored across enterprise applications, cloud platforms and third-party AI services. Maintain an up-to-date inventory to <a href="https://www.techtarget.com/it-strategy/tip/14-tips-for-CIOs-managing-shadow-IT-activities">reduce shadow IT</a>.</li> <li><b>Classify data by risk and business value</b>. Categorize data using a risk-based approach that accounts for sensitivity, intended use, regulatory requirements and business impact. Different AI-generated data requires differing levels of governance.</li> <li><b>Capture provenance and metadata</b>. Record the generating model, prompts or source inputs, creation data, version history, ownership and any subsequent modifications to datasets. The goal is to support transparency, reproducibility and audits.</li> <li><b>Define retention and disposal policies</b>. Determine how long different classifications of AI-generated data should be retained, archived or deleted to meet legal, operational and storage requirements.</li> <li><b>Control access and usage</b>. Apply role-based <a href="https://www.techtarget.com/searchsecurity/tip/CISOs-guide-to-centralized-vs-federated-security-models">access controls</a>, encryption, monitoring and usage policies to prevent unauthorized access or misuse.</li> <li><b>Review, adapt and refine governance</b>. Regularly assess governance policies as AI models, business processes and regulations evolve. Monitor compliance through periodic audits, governance metrics and policy reviews to ensure the framework remains effective and aligned with organizational risk tolerance.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Data governance improves capabilities"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Data governance improves capabilities</h2> <p>Strong governance of AI-generated and synthetic data accelerates innovation rather than restricting it. It also avoids operational and compliance issues that distract the organization from its business objectives.</p> <p>Examples of how data governance speeds up innovation include:</p> <ul class="default-list"> <li>Enabling faster software testing and AI-assisted development.</li> <li>Reducing privacy risks by limiting the use of production data.</li> <li>Implementing <a href="https://www.forbes.com/councils/forbestechcouncil/2026/01/30/governance-by-design-how-to-engineer-trust-in-the-age-of-ai/">governance-by-design</a> rather than governance after deployment.</li> <li>Using AI governance tools to automate lineage, policy enforcement and monitoring.</li> </ul> <p>Measure success with governance KPIs such as provenance coverage, policy compliance rates, audit results and the percentage of AI-generated data under lifecycle management.</p> </section> <section class="section main-article-chapter" data-menu-title="Wrap up: Governance should scale with AI adoption"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Wrap up: Governance should scale with AI adoption</h2> <p>AI-generated and synthetic data must be managed as strategic enterprise assets with both associated value and risk. Organizations that establish clear ownership, risk-based policies and lifecycle governance now will be better prepared for evolving regulations and future technologies while enabling responsible AI innovation.</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to TechTarget Editorial, The New Stack and CompTIA Blogs.</i></p> </section> CIOs need a lifecycle policy for AI generated and synthetic data. Learn how to set lifecycle policies and manage the risks of this type of data. https://cdn.ttgtmedia.com/rms/onlineimages/storage_g1197646065.jpg https://www.techtarget.com/it-infrastructure/tip/Tips-for-the-governance-of-AI-generated-and-synthetic-data Mon, 17 Aug 2026 14:14:00 GMT <p>Constructing a new data center does not begin with architectural plans. It starts with community communication that educates neighbors and opposition leaders.</p> <p>Data center development affects technology, strategy, infrastructure, capital investment, regulatory relationships, reputation and community trust. Communication helps to reduce uncertainty and surface concerns while decisions remain flexible.</p> <p>This article examines why early education matters and what communities need to know about new data centers. It prepares IT leaders to explain the benefits and trade-offs in terms that matter to community members.</p> <section class="section main-article-chapter" data-menu-title="Preparing the message"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Preparing the message</h2> <p>Modern <a href="https://www.techtarget.com/searchDataCenter/feature/Guide-to-understanding-the-various-types-of-data-centers">data centers</a>, especially those centered around AI, are no longer invisible technology infrastructure. Their scale makes issues such as electricity demand, water use, <a href="https://www.techtarget.com/searchdatacenter/feature/The-increasing-concern-of-data-center-land-acquisition">land use</a>, <a href="https://www.techtarget.com/searchdatacenter/tip/Understanding-the-impact-of-data-center-noise-pollution">noise</a>, construction traffic and economic impact highly visible. Different stakeholders may evaluate the same data center project through different lenses.</p> <ul class="default-list"> <li><b>Technology leaders:</b> Capacity, resilience, <a href="https://www.techtarget.com/searchdatacenter/tip/Enhance-operations-with-decentralized-data-centers">performance</a> and security.</li> <li><b>Community leaders:</b> Infrastructure, quality of life, economic benefits and <a href="https://www.techtarget.com/searchDataCenter/tip/Decarbonizing-data-centers-Turning-sustainability-into-strategy">environmental effects</a>.</li> <li><b>Regulators:</b> Compliance, planning, utilities and public interest.</li> </ul> <p>Whether a data center's impact is positive or negative depends on perspective. For IT leaders, this means community opposition is a risk to mitigate.</p> <p>Begin by linking community concerns to executive-level consequences. These include permitting or approval delays, schedule uncertainty, additional mitigation costs and reputational exposure. Risks also include heightened scrutiny from regulators, investors, customers and employees. Proactive communication is essential to risk management and project continuity, not just to reputation management.</p> <p>Delayed communication can allow uncertainty and assumptions to fill the information gap. Early engagement enables organizations to provide a clear, accurate message that can:</p> <ul class="default-list"> <li>Establish fact-based conversations.</li> <li>Surface concerns before project parameters are fixed.</li> <li>Give leaders time to explain technical trade-offs.</li> <li>Identify stakeholders who need specific information.</li> </ul> <p>Communication helps organizations improve preparedness and decision quality. Technology executives need ownership of the underlying facts, trade-offs, concerns and strategic rationale to communicate value accurately.</p> </section> <section class="section main-article-chapter" data-menu-title="The community value proposition"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The community value proposition</h2> <p>IT leaders should articulate the data center's local value proposition, not just its <a href="https://www.techtarget.com/searchdatacenter/tip/Monetizing-modern-data-centers-Strategies-for-IT-leaders">business rationale</a>. Where possible, articulate specific, measurable benefits to enhance credibility. Clearly distinguish projections from commitments.</p> <p>Potential areas to address include:</p> <ul class="default-list"> <li>Direct and indirect employment.</li> <li>Local tax revenue.</li> <li>Infrastructure investment.</li> <li>Local vendors and economic activity.</li> <li><a href="https://www.techtarget.com/searchdatacenter/tip/Preparing-for-the-data-center-talent-shortage">Workforce development</a> and technical training.</li> <li>Potential improvements to regional digital infrastructure.</li> </ul> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The essential message is what the project means for the community, rather than simply why the company needs this facility. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Avoid presenting economic benefits as a reason communities should overlook legitimate concerns.</p> <p>The essential message is what the project means for the community, rather than simply why the company needs this facility.</p> </section> <section class="section main-article-chapter" data-menu-title="Understanding and communicating key technical drivers"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Understanding and communicating key technical drivers</h2> <p>The message must translate technical realities into community and business implications. Concerns typically span several standard areas, and it is up to IT leaders to provide specific information to address them.</p> <h3>Power and grid requirements</h3> <p>Explain that high-density computing, AI workloads and always-on infrastructure can require substantial electrical capacity. Distinguish between facility requirements, existing grid capacity and potential utility or <a href="https://www.techtarget.com/it-infrastructure/news/366643122/Soma-Energy-launches-to-optimize-AI-data-center-energy-use">grid upgrades</a>.</p> <p>Where possible, clarify who is responsible for infrastructure investments.</p> <h3>Water, cooling and environmental considerations</h3> <p>Explain how cooling <a href="https://www.techtarget.com/it-infrastructure/tip/How-do-data-centers-use-and-manage-water">systems impact water</a> and <a href="https://www.techtarget.com/it-infrastructure/tip/How-much-energy-do-data-centers-consume">energy consumption</a>. Note where impacts depend on cooling design, climate, water sources and operating conditions. Avoid presenting one cooling technology as universally superior; cooling approaches will vary by data center.</p> <h3>Land, noise, traffic and physical footprint</h3> <p>Translate specific infrastructure decisions into tangible local impacts. Explain how facility design, equipment placement, landscaping and operating practices can affect the surrounding community. Distinguish between temporary construction and ongoing operational considerations.</p> <h3>Reliability and security</h3> <p>Communicate why redundancy, <a href="https://www.techtarget.com/it-infrastructure/tip/Data-center-power-backup-options-to-deal-with-downtime">backup power</a>, physical security and network resilience add infrastructure and cost. Connect these requirements to the organization's responsibility to maintain critical digital systems.</p> <p>Each technical explanation should answer the question, "Why does this matter to the community?"</p> </section> <section class="section main-article-chapter" data-menu-title="Establishing a communication strategy"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Establishing a communication strategy</h2> <p>Establish a clear communication strategy that emphasizes transparency and prepares the organization for resistance.</p> <h3>Communication strategies</h3> <p>Establishing a clear communication strategy is crucial. Select a communication approach using one of three approaches:</p> <ul class="default-list"> <li><b>Announcement-led.</b> Communicate primarily when a project becomes public. Offers a lower initial effort but greater risk of appearing reactive.</li> <li><b>Information-led.</b> Educate stakeholders about data center infrastructure before specific project decisions. Builds a baseline understanding and reduces unfamiliarity.</li> <li><b>Dialog-led.</b> Combine education with structured opportunities for questions and feedback. Creates opportunities to identify concerns and potentially improve plans.</li> </ul> <h3>Transparency versus uncertainty</h3> <p>Executives may have concerns that early communication will expose plans that are not yet finalized. Leaders can clearly distinguish and explain what is known, what has been evaluated and what remains uncertain. They also identify the factors that drive decisions. Organizations should communicate when additional information will be available.</p> <p>Proactive communication does not require disclosure of confidential or preliminary information.</p> <h3>Preparing for friction</h3> <p>Strong communication will not eliminate all disagreement. Leaders should anticipate scenarios such as misinformation, organized opposition, unexpected infrastructure concerns and changes to project scope. The objective is not to prevent every objection, but to ensure the organization can respond with facts, consistency and credibility.</p> </section> <section class="section main-article-chapter" data-menu-title="Community-centered communication as a strategic capability"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Community-centered communication as a strategic capability</h2> <p>Create a community-centered communication strategy using a three-pronged approach.</p> <h3>Understand stakeholders before developing the message</h3> <p>Identify audiences with different interests and information needs, including residents, local officials, regulators, utilities, businesses, environmental organizations and potential employees. Avoid assuming that one general-purpose message will address every concern.</p> <h3>Establish executive ownership and governance</h3> <p>Coordinate technology, facilities, sustainability, legal, communications, government affairs and executive leaders. Establish a shared source of truth for technical facts, commitments, assumptions and project updates. Define who can make commitments and who is responsible for addressing emerging issues.</p> <h3>Build a two-way information loop</h3> <p>Treat questions and concerns as useful inputs rather than potential hurdles or objections. Categorize issues the organization can address or mitigate versus constraints it cannot change. Listening to a concern does not mean promising the requested outcome.</p> </section> <section class="section main-article-chapter" data-menu-title="The first steps toward communication"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The first steps toward communication</h2> <p>Establish the direction and business objectives before the project narrative. Begin with the organization's broader infrastructure needs and long-term objectives. Explain why resilient computing capacity matters to customers, employees and business continuity.</p> <p>Build a plain-language technical story containing concise explanations of:</p> <ul class="default-list"> <li>Power and grid requirements.</li> <li>Water and cooling.</li> <li>Noise.</li> <li>Traffic and construction.</li> <li>Land use.</li> <li>Economic contribution.</li> <li>Environmental considerations.</li> <li>Reliability and security.</li> </ul> <p>Engage in communication with the community while meaningful decisions remain possible.</p> <ul class="default-list"> <li>Start education before plans are fully finalized.</li> <li>Clearly label assumptions, preliminary plans and firm commitments.</li> <li>Create channels for stakeholder questions and feedback.</li> </ul> <p>Measure understanding and risk by tracking more than communication reach, such as:</p> <ul class="default-list"> <li>Recurring questions.</li> <li>Unresolved concerns.</li> <li>Stakeholder participation.</li> <li>Changes resulting from substantive feedback.</li> <li>Emerging project or reputational risks.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Final thoughts"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Final thoughts</h2> <p>Data center development is more than a technical deployment; it is a community-facing business decision that can face significant opposition. This opposition creates roadblocks, delays and reputational damage. Clear communication begins before the project becomes controversial.</p> <p>CIOs, CTOs and infrastructure leaders should ask:</p> <ul class="default-list"> <li>"What would our community need to understand if we explained this project before asking for its support?"</li> <li>"Which concerns could we address while the project is still flexible?"</li> </ul> <p>A strategic communication approach means educating early, explaining trade-offs honestly, listening before decisions harden and communicating throughout the process.</p> <p><em>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to TechTarget Editorial, The New Stack and CompTIA Blogs.</em></p> </section> Effective communication is essential in data center development. Engaging the community early addresses concerns and prepares IT leaders for stakeholder interactions. https://cdn.ttgtmedia.com/rms/onlineimages/collab_a306983505.jpg https://www.techtarget.com/it-infrastructure/tip/Building-trust-Engaging-communities-in-data-center-development Mon, 17 Aug 2026 12:18:00 GMT <p>IT service management is the set of activities an organization uses to design, build, deliver, operate and support IT services for customers. ITSM is critical to digital transformation, AI adoption and <a href="https://www.techtarget.com/searchitoperations/feature/How-digital-employee-experience-strategy-drives-productivity">rising service expectations</a>. The balance between innovation and operational stability is crucial.</p> <p>IT leaders can't afford service disruptions during modernization processes. However, they also can't allow outdated ITSM offerings to drag on productivity.</p> <p>This article guides IT leaders through a phased approach to modernizing ITSM workflows and measuring their success without disrupting services. Successful ITSM modernization is driven by business outcomes, optimized workflows, phased implementation and measurable improvements -- not wholesale technology replacement.</p> <section class="section main-article-chapter" data-menu-title="Why ITSM modernization fails"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Why ITSM modernization fails</h2> <p>The biggest risks to ITSM modernization initiatives aren't technical; they're organizational. Successful modernization requires executive alignment, governance and a clear business case from the outset.</p> <p>These are the most common pitfalls that derail modernization efforts:</p> <ul class="default-list"> <li>Treating modernization as a platform migration instead of a business transformation initiative.</li> <li>Pursuing "<a href="https://www.techtarget.com/searchCloudComputing/tip/Use-the-7-Rs-to-develop-an-app-modernization-strategy">rip-and-replace</a>" implementations that introduce unnecessary risk.</li> <li>Failing to align ITSM goals with broader business priorities.</li> <li>Lack of executive sponsorship.</li> <li>Lack of cross-functional governance.</li> <li>Poor change management, communication and user adoption.</li> <li>Ignoring legacy process and integration dependencies.</li> </ul> <p>Recognizing these pitfalls early enables organizations to avoid disruptions as they modernize ITSM capabilities.</p> </section> <section class="section main-article-chapter" data-menu-title="Start with workflows, not tools"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Start with workflows, not tools</h2> <p>Modernization doesn't begin with new tools. Instead, it starts with optimizing existing service delivery processes. The focus is on updating current workflows to improve outcomes without adding cost.</p> <p>Use the following steps to evaluate and improve existing workflows:</p> <ul class="default-list"> <li><b>Map and assess existing service workflows. </b>Document current processes to identify inefficiencies, redundancies and opportunities for improvement.</li> <li><b>Identify bottlenecks and repetitive work.</b> Focus on manual tasks and delays that consume time and slow service delivery.</li> <li><b>Standardize high-volume processes before automating. </b>Create consistent workflows to ensure automation improves efficiency. Avoid scaling inconsistencies.</li> <li><b>Improve employee and CX.</b> Design workflows that simplify interactions, reduce wait times and increase satisfaction.</li> <li><b>Evaluate where to enhance existing platforms before replacing them.</b> Maximize current technology investments before committing to costly platform migrations.</li> <li><b>Prioritize integrations that enable business continuity.</b> Strategically connect systems to minimize service disruption during modernization.</li> </ul> <p>Better workflows, not newer software, deliver the greatest return while protecting previous technology investments. Optimized workflows also <a href="https://www.techtarget.com/searchitoperations/tip/Tasks-to-automate-today-to-streamline-IT-operations">enable better automation</a> during the modernization plan phases.</p> </section> <section class="section main-article-chapter" data-menu-title="Modernize in phases, not all at once"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Modernize in phases, not all at once</h2> <p>Reduce operational risk and demonstrate ROI by using a phased modernization approach. The goal is steady progress that avoids disruption.</p> <h3>Phase 1: Assess business priorities and current state</h3> <p>Begin by gathering information on existing business needs and services:</p> <ul class="default-list"> <li>Audit ITSM processes and supporting technologies.</li> <li>Align modernization goals with business objectives.</li> <li>Identify quick wins and define success metrics.</li> </ul> <h3>Phase 2: Optimize and standardize core services</h3> <p>Enable effective automation by standardizing and optimizing core service offerings:</p> <ul class="default-list"> <li>Improve incident, request and change management processes.</li> <li>Remove unnecessary complexity.</li> <li>Standardize workflows before introducing automation.</li> </ul> <h3>Phase 3: Introduce AI and automation strategically</h3> <p><a href="https://www.techtarget.com/searchenterpriseai/tip/Best-practices-for-integrating-third-party-AI-with-local-systems">Integrate AI</a> and automation practices based on the lowest risks and greatest gains:</p> <ul class="default-list"> <li>Automate repetitive, low-risk tasks.</li> <li>Expand self-service capabilities.</li> <li>Pilot AI in targeted use cases before scaling.</li> </ul> <h3>Phase 4: Scale, govern and continuously improve</h3> <p>Establish expectations for continuous improvement and governance:</p> <ul class="default-list"> <li>Expand successful practices across business units.</li> <li>Establish ongoing governance and performance reviews.</li> <li>Refine processes using operational and business metrics.</li> </ul> <p>Phased modernization minimizes disruption, reduces implementation risk and delivers measurable value throughout the journey rather than relying on a single "go-live" push.</p> </section> <section class="section main-article-chapter" data-menu-title="Where AI and automation actually fit"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Where AI and automation actually fit</h2> <p>AI and automation deliver the greatest value when they enhance well-defined ITSM processes rather than replace them. Focus on practical business value, not AI hype.</p> <p>Organizations should begin by <a href="https://www.techtarget.com/searchitoperations/feature/How-runbook-automation-reduces-IT-operational-costs">automating repetitive, rules-based tasks</a>. Find quick wins that reduce manual effort, improve response times and enable IT staff to focus on higher-value initiatives. Here are some examples:</p> <ul class="default-list"> <li>Intelligent ticket routing.</li> <li>Self-service virtual agents.</li> <li>Knowledge recommendations.</li> <li>Predictive incident detection.</li> <li>Workflow orchestration.</li> </ul> <p>AI further improves ITSM by recommending technical resources, powering self-service virtual agents, predicting potential incidents and identifying patterns that help prevent recurring issues. However, these capabilities are most effective when supported by standardized workflows and high-quality data. Automating inconsistent or poorly designed processes simply accelerates inefficiency.</p> <p>IT leaders should also establish clear governance for responsible AI adoption aligned with organizational policies. <a href="https://www.techtarget.com/searchDataCenter/tip/AI-operating-models-Balancing-autonomy-and-human-oversight">Human oversight</a> remains essential for major incidents, complex change management and decisions that carry significant business or security risks. AI should augment the expertise of IT professionals, not replace it.</p> </section> <section class="section main-article-chapter" data-menu-title="Measure success without disrupting service"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Measure success without disrupting service</h2> <p>Tie modernization success to business outcomes and operational improvements. These measures track actual improvement and ROI.</p> <p>Operational metrics:</p> <ul class="default-list"> <li><a href="https://www.techtarget.com/searchDisasterRecovery/feature/Disaster-recovery-glossary-Terms-to-know">Mean time to resolution</a>.</li> <li>Change success rate.</li> <li>Service-level agreement compliance.</li> <li>Incident volume and backlog.</li> <li>Automation rate.</li> <li>Self-service adoption.</li> </ul> <p>Business metrics:</p> <ul class="default-list"> <li><a href="https://www.techtarget.com/searchitoperations/tip/How-DEX-metrics-help-build-a-better-digital-workplace">Employee satisfaction</a>.</li> <li>Customer satisfaction.</li> <li>Cost per service request.</li> <li><a href="https://www.techtarget.com/searchdisasterrecovery/feature/Disaster-recovery-for-small-businesses-leaves-no-room-for-excuses">Reduced downtime</a>.</li> <li>Reduced time to deliver new business services.</li> <li>IT operating efficiency.</li> </ul> <p>The strongest modernization programs demonstrate improvements in resilience, efficiency and business performance.</p> </section> <section class="section main-article-chapter" data-menu-title="Conclusion"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Conclusion</h2> <p>ITSM modernization isn't a one-time project with a defined finish line. As business needs, technologies and customer expectations continue to evolve, organizations must continually refine how they deliver IT services. Viewing modernization as an ongoing capability, not simply a technology upgrade, helps IT remain responsive and resilient.</p> <p>The strongest modernization efforts start with the business, not the technology. Align initiatives with organizational goals, improve workflows before investing in new tools, introduce changes in phases and apply AI where it adds real value. IT leaders can use these approaches to make meaningful progress without disrupting day-to-day operations.</p> <p>A measured, governance-driven <a target="_blank" href="https://www.freshworks.com/it-service-management-software/itsm-governance/" rel="noopener">approach</a> reduces risk while building momentum and showcasing ROI. Instead of chasing large-scale transformation, many organizations achieve lasting improvements that strengthen service quality, increase operational agility and position IT as a strategic partner in long-term business growth. Focus on steady progress that strengthens both operations and the organization.</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to Informa TechTarget, The New Stack and CompTIA Blogs.</i></p> </section> ITSM modernization succeeds by optimizing workflows, implementing changes in phases, aligning with business goals and applying AI to improve services with minimal disruption. https://cdn.ttgtmedia.com/rms/onlineimages/disaster_recovery_a379640336.jpg https://www.techtarget.com/it-infrastructure/tip/How-to-modernize-ITSM-without-disrupting-operations Mon, 17 Aug 2026 10:51:00 GMT <p>Autonomous infrastructure comprises AI-driven, self-managing policy-based systems for IT operations. It shifts from rigid automation to data-driven decision-making. Modern IT ops require increasing AI autonomy to achieve the best results. The critical intersection is:</p> <ul class="default-list"> <li>Complexity of hybrid and multi-cloud environments.</li> <li>Advances in AI and observability.</li> <li>Pressure to reduce costs and improve resilience while maintaining agility.</li> </ul> <p>IT professionals now act as governors, not just operators. However, new risks in security, data and accountability must be actively managed. Organizations that invest in governance, data quality and workforce evolution will unlock their full value while avoiding systemic risk.</p> <section class="section main-article-chapter" data-menu-title="Automation value and risk"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Automation value and risk</h2> <p>For IT leaders, balancing value against risk now means shifting some risks from human control to system autonomy.</p> <p>Where autonomy delivers value:</p> <ul class="default-list"> <li><b>Availability and resilience</b>. Reduce downtime using <a href="https://www.techtarget.com/searchitoperations/tip/Self-healing-IT-systems-Benefits-challenges-and-use-cases">self-healing systems</a> and predictive remediation.</li> <li><b>Cost optimization.</b> Enable dynamic resource allocation and scaling with less manual intervention.</li> <li><b>Operational efficiency.</b> Provide lower incident volume, faster resolution and reduced administrative effort.</li> </ul> <p>Where risk emerges:</p> <ul class="default-list"> <li><b>Decision transparency.</b> Limited explainability of AI-driven actions and challenges in root cause analysis.</li> <li><b>Failure propagation at scale.</b> Misconfigured policies or faulty models can cascade rapidly.</li> <li><b>Security implications.</b> AI control planes are high-value targets. Autonomous responses may unintentionally disrupt services and expose them to adversarial inputs or model manipulation.</li> <li><b>Data dependency and quality.</b> Watch reliance on accurate telemetry and historical data, as well as biased or incomplete data that can drive flawed decisions.</li> <li><b>Vendor lock-in and interoperability constraints.</b> Proprietary orchestration platforms and difficulty integrating across legacy and multi-cloud environments.</li> </ul> <p>Autonomous infrastructure delivers measurable gains but changes the risk catalog. New governance practices are required to protect operations.</p> </section> <section class="section main-article-chapter" data-menu-title="Governance enables safe autonomy"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Governance enables safe autonomy</h2> <p>Governance is critical for autonomous infrastructure. The organization is shifting from a traditional manual or automated operational approach to a policy-driven control system that relies on <a target="_blank" href="https://cloudsecurityalliance.org/blog/2026/01/28/levels-of-autonomy" rel="noopener">AI-based autonomy</a>. Governance ensures this autonomy enhances, not replaces, accountability. Some core governance principles include:</p> <h3>Human-in-the-loop oversight</h3> <ul class="default-list"> <li>Define autonomy tiers, including advisory, supervised and fully autonomous. Not every system requires full autonomy, especially when risk increases.</li> <li>Require approval for high-impact actions and novel scenarios, allowing human decision-making to override autonomy.</li> </ul> <h3>Auditability and observability</h3> <ul class="default-list"> <li>Establish full traceability of decisions, inputs and actions to foster trust and prove compliance.</li> <li>Configuring logging for compliance, forensics and continuous improvement.</li> </ul> <h3>Policy-as-code framework</h3> <ul class="default-list"> <li>Encode business intent, including cost thresholds, performance goals and risk tolerance.</li> <li>Align system behavior with enterprise priorities.</li> </ul> <p>AI-driven controls require additional management policies to meet compliance goals and reduce risk. Governance must implement these policies by:</p> <ul class="default-list"> <li>Establishing clear ownership of autonomous decisions.</li> <li>Aligning autonomy with enterprise AI ethics standards.</li> <li>Constructing guardrails to block unintended or non-compliant actions.</li> </ul> <p>Technical control mechanisms include "circuit breakers" to halt unsafe behavior before it disrupts operations or cascades across systems. Sandboxed testing and simulation before deployment also reduces unintended consequences and demonstrates due diligence.</p> <p>Governance directives will define escalation paths for anomalies, enabling humans to enter the loop and evaluate potentially disruptive behavior. Governance is the scaling mechanism for trust. Organizations that formalize it early will adopt autonomy faster, more safely and with better compliance.</p> </section> <section class="section main-article-chapter" data-menu-title="Redefining IT ops"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Redefining IT ops</h2> <p>Evolving an enterprise environment to an autonomous infrastructure requires changing roles and updating measurement practices.</p> <p><b>Changing IT roles and skills</b></p> <ul class="default-list"> <li>Shift from manual operators or automation authors to policy designers, automation architects and system governors.</li> <li>Increased demand for hybrid expertise consisting of infrastructure, data management, AI and governance skills.</li> </ul> <p><b>Changing metrics</b></p> <ul class="default-list"> <li>Move away from measuring ticket volume and mean time to resolution.</li> <li>Opt for autonomous resolution rate, policy effectiveness, and business-aligned reliability outcomes.</li> </ul> <p>The primary transformation is not technical; it's organizational and cultural. It includes relying on cross-functional DevOps, <a href="https://www.techtarget.com/searchsecurity/tip/Streamline-SecOps-with-SOAR-workflows-and-playbooks">SecOps</a> and <a href="https://www.techtarget.com/searchcloudcomputing/tip/The-business-case-for-FinOps-to-control-cloud-spending">FinOps</a> teams. Organizations must foster trust in autonomous systems and satisfy resistance to role changes across teams. Allocate resources for change management and employee <a href="https://www.techtarget.com/searchCloudComputing/tip/How-to-use-AI-to-bridge-the-cloud-skills-gap">upskilling</a>.</p> </section> <section class="section main-article-chapter" data-menu-title="Strategic implications and adoption path"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Strategic implications and adoption path</h2> <p>Realizing the competitive advantages offered by autonomous infrastructures will depend on how quickly and safely it is adopted. Benefits depend on safe, guided adoption. Poor implementation raises risks; good governance increases rewards.</p> <p>Carefully consider and allocate resources for the following economic and investment realities while simultaneously recognizing benefits:</p> <ul class="default-list"> <li><b>Upfront costs</b>. Platforms, integration and training.</li> <li><b>ROI drivers. </b>Reduced downtime, labor efficiency and optimized infrastructure spend.</li> <li><b>Trade-offs.</b> Prioritize build versus buy decisions for autonomous systems, as well as flexibility versus vendor dependence decisions.</li> </ul> <h3>Adoption roadmap</h3> <p>Begin with low-risk, high-value domains such as observability, incident triage or cost optimization, where outcomes are measurable and reversible. Establish clean data pipelines and a unified telemetry platform.</p> <p>Progress through the following three maturity stages:</p> <ol class="default-list"> <li><b>Assisted automation.</b> Human-led, AI-supported.</li> <li><b>Semi-autonomous.</b> Policy-bound execution with oversight.</li> <li><b>Autonomous domains.</b> Fully self-managing within guardrails.</li> </ol> <p>Companies should establish governance and policy controls early. Also, use pilot programs to build trust, validate ROI and refine escalation models before attempting a broader rollout.</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to TechTarget Editorial, The New Stack and CompTIA Blogs.</i></p> </section> Autonomous infrastructure is not just implementing new technology; it completely redefines IT's role. Success depends on balancing speed, safety and accountability. https://cdn.ttgtmedia.com/rms/onlineimages/disaster_recovery_g175031469.jpg https://www.techtarget.com/it-infrastructure/tip/How-autonomous-infrastructure-alters-IT-operations Mon, 17 Aug 2026 10:40:00 GMT <p>In recent years, the AI boom has completely changed data center economics. While organizations naturally gravitate toward managing <a href="https://www.techtarget.com/searchdatacenter/tip/How-much-energy-do-data-centers-consume">energy consumption</a> costs tied to compute, storage can also be a significant contributor to workload TCO. Organizations that treat energy efficiency as a primary goal when designing their storage architecture may be able to lower operating costs, increase rack densities and more easily satisfy the organization's sustainability objectives. In some cases, aggressively managing storage power consumption now might even put off expensive facility upgrades for a time.</p> <section class="section main-article-chapter" data-menu-title="How AI workloads are driving data center power concerns"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How AI workloads are driving data center power concerns</h2> <p>Historically, one of the biggest challenges for data center operators has been accommodating all of an organization's IT resources without running out of floor space. After all, nobody wants to shoulder the cost of a data center expansion or the construction of a new facility. Today, however, the availability of sufficient <a href="https://www.techtarget.com/data-technologies/feature/Power-constrained-data-architecture-curbing-AI-ambitions">electricity is becoming a bigger operational constraint</a> than floor space.</p> <p>AI servers consume far more power than the servers used to host traditional workloads. According to <a href="https://www.socomec.us/en-us/solutions/business/data-centers/understanding-power-consumption-data-centers">some estimates</a>, "Traditional data centers operate at 5-10 kW per rack, while AI-optimized facilities now require 60+ kW per rack within the same square foot footprint." Most of the datacenters built before the AI boom were never designed to accommodate such demand, and utility expansions can take years to complete.</p> <p>Simply put, power has gone from being an operating expense to becoming a business constraint. Hence, organizations must take immediate steps to ensure that they are using the available power efficiently. Every watt of power that is being consumed by storage resources is a watt that cannot be allocated to the AI infrastructure.</p> </section> <section class="section main-article-chapter" data-menu-title="What Does Storage Actually Contribute to Your data center Power Bill?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What Does Storage Actually Contribute to Your data center Power Bill?</h2> <p>Like any piece of data center hardware, storage arrays consume power. Even a relatively small storage appliance can easily consume over 100 watts. Of course, larger arrays consume more power. According to <a href="https://www.solved.scality.com/data-center-power-efficiency/">Solved Magazine</a>, a 2U SAN might contain 20 to 40 drives and consume 2 to 5 KW of power. As an example, the Dell PowerMax 8500's power consumption varies based on its configuration. However, the smallest number <a href="(https:/www.delltechnologies.com/asset/en-us/products/storage/technical-support/powermax-2500-8500-spec-sheet.pdf">given by Dell</a> is 7.124 kVA (approximately 5.6 KW).</p> <p>If these numbers sound high, it's worth remembering that in an enterprise-class environment, storage hardware does far more than just hosting drives. Functions such as erasure coding and encryption consume significant compute resources, which in turn increase power consumption.</p> <p>It's also worth considering that the storage hardware itself is only half of the story. As the storage hardware consumes power, it releases heat as a byproduct. This thermal load forces cooling systems to work harder, further increasing power consumption. In other words, every watt consumed becomes heat that has to be removed.</p> </section> <section class="section main-article-chapter" data-menu-title="Intelligent Tiering as an Energy Strategy"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Intelligent Tiering as an Energy Strategy</h2> <p>AI workloads generally require access to vast inference, analytics or training datasets amounting to many terabytes (TB) or even petabytes, so the need for storage hardware certainly isn't going away. Even so, it may be possible to drive down costs through <a href="https://www.techtarget.com/it-infrastructure/tip/Why-storage-tiering-is-necessary-now-more-than-ever">intelligent storage tiering</a>.</p> <p>Storage tiering was introduced at a time when flash storage offered extremely limited capacity with a very high cost per gigabyte. The idea was that not all data requires the same level of performance. Hence, frequently accessed "hot" data could be placed on high-performance SSDs, while occasionally used "warm" data could reside on less expensive, higher-capacity spinning media. Rarely accessed "cold" data could be moved to cheap archival disks or even kept offline.</p> <p>As flash storage prices began to decline, many organizations moved their performance-sensitive workloads to all-flash arrays, with traditional HDD/SSD tiering being used less frequently for primary workloads. However, intelligent tiering is beginning to see a resurgence thanks to the need for reducing power consumption in the data center.</p> <p>Simply put, higher-performance storage tends to use more power, and because it is consuming more power, it also produces more heat. This is not necessarily to say that an SSD uses more power than a spinning disk. In some cases, an SSD might actually use a bit less power than spinning media.</p> <p>What needs to be considered, however, are metrics such as watts per IOPS, watts per transaction, watts per TB, or watts per workload. An all-flash array might have no trouble accommodating cooler data. Still, if power efficiency is a priority, then that data might be better suited to higher-capacity HDD-based storage or archival tiers -- some of which can reduce power consumption by spinning down inactive drives.</p> <p>This means that rather than placing an entire data set on high-performance storage, it may be more efficient to reserve the flash media for the data that is being actively used. Typically, only a fraction of the overall data is in use at any given moment, meaning that cooler data can be stored on slower but more energy-efficient tiers. And if some of the less frequently used data suddenly becomes important, it can be automatically moved to the high-performance tier, where it will remain until it "cools off."</p> <p>Storage tiering can also sometimes reduce hardware acquisition costs by ensuring only frequently used data occupies expensive high-performance storage. It can also reduce power costs by keeping infrequently used data on lower-power, lower-cost storage that requires less energy and cooling. This, in turn, ensures that less energy is being consumed by storage, thereby making that excess energy available for use with AI workloads.</p> </section> <section class="section main-article-chapter" data-menu-title="Balancing ESG Goals with Storage Needs"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Balancing ESG Goals with Storage Needs</h2> <p>Organizations that design their storage architecture for operational efficiency might discover that their efforts naturally support ESG reporting, as reduced power consumption is objectively measurable and easy to document. In the case of an AI data center, however, the business justification for reducing power consumption might have less to do with meeting ESG goals and more to do with preventing a limited supply of power (or a limited power and cooling budget) from becoming an operational constraint.</p> <p>In the AI era, those responsible for storage must look beyond performance and capacity planning. Today, storage decisions play a pivotal role in reducing an organization's <a href="https://www.techtarget.com/searchdatacenter/definition/TCO">TCO</a> while also ensuring that limited data center power supplies can be used where they will deliver the greatest business value.</p> <p><i>Brien Posey is a former 22-time Microsoft MVP and a commercial astronaut candidate. In his more than 30 years in IT, he has served as a lead network engineer for the U.S. Department of Defense and a network administrator for some of the largest insurance companies in America.</i></p> </section> AI has turned power into a bottleneck, making energy-efficient storage design and intelligent tiering central to cutting down on watts and freeing capacity for AI. https://cdn.ttgtmedia.com/rms/onlineimages/disaster_recovery_a257795847.jpg https://www.techtarget.com/it-infrastructure/tip/Energy-Efficient-Storage-Architecture-in-the-AI-Era Mon, 17 Aug 2026 10:39:00 GMT <p>Matt Pocock’s skills remained No. 1 for the week of Aug. 9–16 with 8,231 stars gained. &nbsp;</p> <p>orca entered at No. 2, while agency-agents reached No. 3 and pi climbed from No. 10 to No. 5. &nbsp;</p> <p><a target="_blank" href="https://github.com/public-apis/public-apis" rel="noopener">Public-apis</a> made a notable jump to No. 4, while Agent-Reach fell from No. 4 to No. 13 and taste-skill dropped from No. 8 to No. 18. &nbsp;</p> <p>AI-For-Beginners, graphify, build-your-own-x and generative-ai-for-beginners were among the repositories that dropped out of the top 20.&nbsp;</p> <section class="section main-article-chapter" data-menu-title="Top 20 GitHub repos this week"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Top 20 GitHub repos this week</h2> <p>GitHub users star a project when they want to bookmark it, like it or show appreciation to the creator of the repo. In this way, stars act as a proxy for developer community interest. While any attention-based ranking mechanism is prone to manipulation, IT leaders can still make use of this information. GitHub star growth can signal emerging tools, or tools that teams might ask for next. Star growth metrics also give a window into developer demand, highlighting shadow IT that developers might be experimenting with. Experimenting with <a href="https://www.darkreading.com/application-security/openclaw-ai-runs-wild-business-environments">fast-growing repos can introduce security vulnerabilities</a>, so knowledge of repos in-high demand can help direct the attention of security teams. The top repo list, in aggregate, can also show what <i>types </i>of projects overall are gaining traction in the developer community.</p> <p><iframe title="GitHub weekly trending repositories" aria-label="Table" id="datawrapper-chart-Z6fmP" src="https://datawrapper.dwcdn.net/Z6fmP/3/" scrolling="no" frameborder="0" style="width: 0; min-width: 100% !important; border: none;" height="935" data-external="1"></iframe></p> <p> <script type="text/javascript">window.addEventListener("message",function(a){if(void 0!==a.data["datawrapper-height"]){var e=document.querySelectorAll("iframe");for(var t in a.data["datawrapper-height"])for(var r,i=0;r=e[i];i++)if(r.contentWindow===a.source){var d=a.data["datawrapper-height"][t]+"px";r.style.height=d}}});</script> </p> <p><i>Ben Lutkevich is an award-winning technology writer and editor.</i></p> </section> GitHub Stars are a proxy for developer interest. Weekly GitHub star growth highlights fast-rising repos, giving early insight into adoption trends, risks and strategic opportunities. https://cdn.ttgtmedia.com/visuals/digdeeper/4.jpg https://www.techtarget.com/it-infrastructure/tip/What-repos-are-trending-on-GitHub Mon, 17 Aug 2026 09:15:00 GMT <p>Enterprise AI is entering a new phase with competitive advantage shifting from better responses to reliably completing business tasks.</p> <p>The next generation of AI will be judged by how reliably it completes work, not by how 'intelligently' it responds.</p> <p>The basic unit of work has evolved from code to prompts to execution loops. This article explains why prompts alone reach a practical limit. It then covers execution loops, why they matter and how they change software development and governance. Finally, it provides IT leaders with the next steps to take to evolve beyond prompt engineering.</p> <section class="section main-article-chapter" data-menu-title="Enterprise AI is entering a new phase"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Enterprise AI is entering a new phase</h2> <p>The first wave of enterprise-level AI centered on copilots, chatbots and prompt engineering. As foundation models mature, actual capability is becoming more crucial to practical implementations. Executive priorities are shifting toward operational efficiency, measurable ROI and workflow automation. Governance and reliability are increasingly crucial as enterprise AI takes over essential tasks and workflows.</p> <p>So why aren't better prompts enough anymore? Because enterprises need to measure success by completed work, not well-written responses. That shift changes the fundamental unit of work from prompts to execution loops.</p> </section> <section class="section main-article-chapter" data-menu-title="Why loops matter more than better prompts"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Why loops matter more than better prompts</h2> <p>Traditional software uses code that explicitly defines every step of a process. Generative AI adds a new abstraction: Prompts that enable users to describe what they want in natural language. That shift drastically lowered the barrier to interacting with software. However, a prompt typically produces a response rather than accomplishing a task.</p> <p>Execution loops represent the next stage in that evolution. Rather than simply generating an answer, an agent operating within a loop can plan the work, execute actions, verify results, adapt when something goes wrong and continue until the objective is complete or requires human intervention.</p> <p>Loops don't replace code or prompts; enterprise AI still depends on both. Code provides the rules, integrations and guardrails. Prompts help the model reason about a task. Execution loops orchestrate the process, coordinating planning, action, verification and recovery. This shift from generating responses to reliably delivering outcomes defines the difference between an AI assistant and an AI agent.</p> <p>Prompts typically:</p> <ul class="default-list"> <li>Generate a response.</li> <li>Assume success.</li> <li>End after one interaction.</li> </ul> <p>Execution loops typically:</p> <ol class="default-list"> <li>Plan work.</li> <li>Execute actions.</li> <li>Observe results.</li> <li>Verify outcomes.</li> <li>Retry failures.</li> <li>Escalate exceptions.</li> <li>Stop only when objectives are complete.</li> </ol> <p>Consider the simple analogy of a GPS device. It doesn't just provide one direction; it continually checks progress and reroutes until the user reaches the destination.</p> <p>In a basic enterprise example, a prompt assistant might generate deployment instructions for an IT tech to use. An agent loop would verify permissions, create infrastructure, validate the deployment, retry failed steps, open tickets if needed and produce an audit record.</p> </section> <section class="section main-article-chapter" data-menu-title="From assistants to operational agents"> <h2 class="section-title"><i class="icon" data-icon="1"></i>From assistants to operational agents</h2> <p>As AI capabilities evolve, it's crucial to distinguish between systems that assist people and those that perform work. Prompt-based assistants excel at generating content, answering questions and offering recommendations. Operational agents build on these capabilities by executing tasks, monitoring progress and adapting until an objective is achieved.</p> <h3>A practical framework for executives</h3> <p>A useful way to think about the difference is this: Assistants generate responses; agents deliver outcomes. An assistant might explain how to provision a new development environment, while an operational agent can request approvals, create the necessary resources, verify the deployment and escalate issues if something goes wrong. The defining characteristic isn't a more powerful model; it's the execution loop that enables the system to plan, verify and recover.</p> <h3>Execution loops operate at different scales</h3> <p>Enterprises can use execution loops at multiple levels to complete specific outcomes.</p> <ul class="default-list"> <li><b>Task loops.</b> These complete a single objective, such as resetting a password or approving an expense report.</li> <li><b>Workflow loops.</b> These coordinate multiple tasks across applications and teams.</li> <li><b>System loops.</b> These oversee multiple agents, enforcing policies, managing shared resources and maintaining operational health.</li> </ul> <h3>Developers increasingly design execution systems</h3> <p>Rather than replacing developers, execution loops show where <a href="https://www.techtarget.com/searchCIO/tip/AI-augmented-teams-Training-for-human-machine-collaboration">engineering effort creates value</a>. Developers increasingly focus on defining business objectives, connecting enterprise systems, establishing verification logic and building observability. Instead of writing the best prompt, they ask, "How does the system do the work and know it is complete?"</p> </section> <section class="section main-article-chapter" data-menu-title="Governance, reliability and observability"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Governance, reliability and observability</h2> <p>As AI systems take on operational responsibilities, reliability matters more than intelligence alone. An impressive response has little business value if the work is left incomplete or errors go undetected. Execution loops improve reliability by verifying results, recovering from failures and escalating exceptions when needed. They don't eliminate mistakes, but they make failures visible, manageable and far less likely to propagate unnoticed.</p> <p>Where a prompt might summarize suspicious login activity, a loop-based agent correlates alerts, gathers additional information, opens an investigation and escalates the situation to an analyst once predefined thresholds are met.</p> <p>Governance includes human approval for high-risk actions, identity and permissions controls, audit trails, bonded execution and <a href="https://www.techtarget.com/ai/tip/AI-FinOps-requires-a-different-approach-than-cloud">cost controls</a>. Human oversight shifts from reviewing every action to supervising exceptions and high-impact decisions.</p> <p>Enterprise AI should automate routine work while keeping <a href="https://www.techtarget.com/searchdatacenter/tip/Balancing-automation-with-human-oversight-in-AI-data-centers">people responsible for high-impact decisions</a>. Execution loops might access enterprise systems, make repeated decisions and consume resources. Teams must set clear boundaries for when agents can act independently versus when they must request human approval. This approach lets teams scale automation without sacrificing accountability, compliance or security.</p> <p>Continuous governance requires continuous visibility. IT leaders should be able to answer straightforward questions about AI agents, such as:</p> <ul class="default-list"> <li>What actions did the agent take?</li> <li>Why did it make those decisions?</li> <li>What happened when something failed?</li> <li>Can the entire process be reconstructed for an audit?</li> </ul> <p>These policies, along with policy enforcement, identity controls and cost monitoring, are central to evaluating agentic AI platforms. The goal isn't simply deploying more autonomous systems, but deploying systems whose behavior is transparent, trustworthy and aligned with enterprise governance requirements.</p> </section> <section class="section main-article-chapter" data-menu-title="What leaders should do next"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What leaders should do next</h2> <p>Execution loops increase both capability and responsibility, delivering strategic benefits and trade-offs.</p> <p>Benefits:</p> <ul class="default-list"> <li>More reliable execution.</li> <li>Measurable business outcomes.</li> <li><a href="https://www.techtarget.com/data-technologies/tip/Resilience-strategies-for-an-AI-powered-era">Operational resilience</a>.</li> <li>Greater automation.</li> </ul> <p>Trade-offs:</p> <ul class="default-list"> <li>More integration effort.</li> <li>Increased governance requirements.</li> <li>Greater operational complexity.</li> </ul> <p>IT leaders should evaluate AI platforms based on execution, recovery, governance and observability, not model benchmarks alone.</p> <h3>Next steps</h3> <p>Begin with repeatable, well-defined operational workflows where success can be clearly measured. Evaluate vendors by asking:</p> <ul class="default-list"> <li>How does the system verify success?</li> <li>How does the system recover from failure?</li> <li>Where are humans involved?</li> <li>What audit and observability capabilities are built in?</li> </ul> <p>Measure success by using business outcomes rather than response quality alone. Prompt engineering enabled enterprises to experiment with generative AI. Execution loops will determine whether AI becomes a dependable operational infrastructure.</p> <p>Organizations that lead the next phase of AI adoption will likely be those that design systems capable of planning, verifying, adapting and governing work, not simply generating increasingly sophisticated responses.</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to TechTarget Editorial, The New Stack and CompTIA Blogs.</i></p> </section> Execution loops move enterprise AI beyond prompt engineering, enabling agents to plan, act, verify results, recover from failures and deliver outcomes. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1133924836.jpg https://www.techtarget.com/it-infrastructure/tip/Prompts-vs-loops-Why-prompt-engineering-wont-build-your-AI-workforce Fri, 14 Aug 2026 15:52:00 GMT <p>Hardened images are becoming a practical necessity for many organizations running production workloads.</p> <p>Production workloads incorporate third-party software that enterprise development teams have no control over. Generic containers hold many components that might not be relevant to a given use, but that still introduce a certain level of risk if the software components they contain are compromised. If someone finds a vulnerability in those upstream sources, engineering teams have no direct way to fix it on their own.</p> <p>Vendors, including BellSoft, Chainguard, Docker, Minimus, Red Hat, RapidFort and Wiz, aim to prevent this issue by providing container images with vulnerabilities already resolved, called hardened images.</p> <p>In a recent <a href="https://research.esg-global.com/aim/en/reports/515202223">Omdia survey</a>, 77% of respondents said that either third-party and OSS code or software dependencies were the riskiest elements of the software supply chain for cloud-native technology. The same survey found that hardened container images were the only security tools for third-party and OSS code components that more than half of organizations identified as very effective.</p> <section class="section main-article-chapter" data-menu-title="Why are hardened container images important now?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Why are hardened container images important now?</h2> <p>Hardened images are not new, but they have become more relevant due to a confluence of factors.</p> <p>Software supply chains have become more complex, attacks on them have become more common and drawn more attention, and vulnerability discovery has increased in quantity and speed -- in part because of AI. In addition, organizations have historically been focused on <i>removing </i>friction for developers, not adding it with additional security measures.</p> <p>"It still persists, the idea that you don't create friction in how developers work. They didn't get a lot of security historically" said Katie Norton, analyst at IDC. "I'm meeting more and more companies that when they think about securing AI, the developer use case around that is so prominent because of things like Claude and Cursor and AI coding. The software development process is considered use case number one, just as much as Kathy in HR using ChatGPT when she's not supposed to. The use cases around coding with AI are just so visible that now the priority level has just jumped [several] notches."</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">What are the features of hardened container images?</h3> <p>The features of hardened images generally include:</p> <ul class="default-list"> <li>A base image following security standards that map closely to NIST SP 800-190's container security guidance.</li> <li>A formal record of the components included in the image and supply-chain relationships of the components -- a software bill of materials.</li> <li>Cryptographic provenance ensuring that the image was not tampered with.</li> <li>A commitment from the image provider or vendor to minimize exposure to vulnerabilities, and to remediate and republish images quickly when upstream vulnerabilities are discovered.</li> </ul> </div> </div> </section> <section class="section main-article-chapter" data-menu-title="Benefits of hardened container images"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Benefits of hardened container images</h2> <p>There are a confluence of factors driving organizations to realize the benefits of hardened images.&nbsp;</p> <h3>1. Software supply chain attacks and zero-trust security</h3> <p>Containers were not initially designed with security in mind; they were designed for convenience. They traditionally came filled with bloat -- packages, binaries and libraries that a particular application will never actually use. <br><br>"The images that Minimus, Chainguard and everybody else are providing are so stripped down, there's not much in there for attackers to use to live off the land. So, there is this real reduction in post-compromise risk and post-compromise attack surface," said Neil Carpenter, Go-to-market at Minimus, a vendor in the vulnerability-free container space.</p> <h3>2. Compliance</h3> <p>Regulatory requirements, including <a href="https://www.datacenterknowledge.com/cloud/what-is-fedramp-federal-risk-and-authorization-management-program">FedRAMP</a>, the EU Cyber Resilience Act and <a href="https://www.techtarget.com/whatis/definition/What-is-the-Digital-Operational-Resilience-Act-DORA">DORA</a>, have made hardened container images necessary.</p> <p>"FedRAMP is probably the deepest end of the vulnerability management pool," Carpenter told TechTarget. "For FedRAMP, you have to report monthly on every vulnerability in your environment and what your plan for addressing it is. Which is insane … being able to replace every public image you're using -- [containing] hundreds of vulnerabilities -- with ones that are tightly managed is a huge advantage for those shops."</p> <h3>3. Removing toil for developers</h3> <p>It's onerous for developers to manually monitor, maintain and patch the hundreds of libraries that come packaged in container images. That's where hardened container images come in. They strip images down to limit the attack surface.</p> <p>"There's a real advantage in making vulnerability management work easier," Carpenter said.</p> <p>He described his previous experience at Orca, where its security product gave dev teams instant security visibility. This service often surfaced hundreds of thousands of vulnerabilities in 15 minutes. This puts development teams tasked with shifting security left and applying app sec at the build stages in sticky situations.</p> <p>"Now there's a ton of friction and the first thing the developer is trying to do is figure out how to route around security because they have real goals around shipping features, fixing bugs, etc.," Carpenter said.&nbsp;</p> <p>Starting with a hardened image and builder ensures developers only need to worry about the vulnerabilities a developer introduced, instead of the thousands of vulnerabilities that come from off-the-shelf public images.</p> <h3>4. Prevention early in the development cycle</h3> <p>Instead of scanning for and detecting vulnerabilities downstream, hardened container images focus on preventing them from the start.</p> <p>The move to prevention from detection comes as the signals that vulnerability scanners rely on are becoming less dependable.</p> <p>"Going into 2026, the combination of gov't funding issues, layoffs at NIST and CISA, rising numbers of legit CVEs, and the AI-driven CVE slop means that the data we need for vulnerability management is going to get worse rather than better," wrote Carpenter in a November 2025 LinkedIn <a href="https://www.linkedin.com/posts/neil-carpenter_the-more-i-think-about-jerry-gamblins-aspirational-share-7398015266716012544-R-yA">post</a>.</p> <p>NIST announced in April 2026 that it would reduce the amount of CVEs that <a href="https://www.linkedin.com/pulse/missing-dhi-vulnerabilities-neil-carpenter-ikdje/">get enriched</a> with data that scanners rely on. This means it is <a href="https://www.darkreading.com/cybersecurity-operations/can-global-decentralized-system-save-cve-data">no longer a dependable single source of truth</a>, creating problems for scanners that count on it as a lone signal. Carpenter covers the changing, more decentralized landscape of NVD enrichment in a Minimus blog post from June.</p> <p>Preventing risk at the beginning of development narrows the scope of vulnerabilities on scanners are responsible for later in the SDLC.</p> <h3>5. Performance</h3> <p>Hardened images also offer performance improvements. Generic container images are a lot larger than normal images, containing vestigial packages, binaries and libraries that might never be needed. More bloatware means more potential performance problems.</p> <p>"So, if we compare base images, just check the size. The size may be different by multiple times -- like 100 megabytes versus 400 or 500 megabytes -- and that's the image to run the same application, which means 75% of it is bloatware," said Dmitry Chuyko, senior performance architect at BellSoft.</p> <p><i>Ben Lutkevich is an award-winning technology writer and editor covering IT infrastructure, app development and AI.</i></p> </section> Hardened container images reduce software supply chain risk by minimizing vulnerabilities, improving compliance, reducing toil and boosting performance. https://cdn.ttgtmedia.com/rms/onlineimages/code_g136298313.jpg https://www.techtarget.com/it-infrastructure/tip/5-reasons-hardened-images-are-seeing-a-resurgence Fri, 14 Aug 2026 15:34:00 GMT <p>IT service management is being transformed by AI, as ITSM software providers add AI capabilities to support end-to-end IT service delivery management.</p> <p>Although <a href="https://www.techtarget.com/searcherp/feature/How-to-decide-which-AI-features-are-worth-turning-on">increasing AI capabilities</a> garner a lot of attention, ITSM is also evolving in other ways in its quest to support IT departments in designing, delivering, managing and improving their services.</p> <p>As research firm Gartner noted in its most recent market guide for ITSM platforms, "many ITSM platform providers focus their product roadmaps on emerging features, including AI and non-IT case workflows, rather than building more robust core features such as reporting, automation, integration and product usability."</p> <p>Despite the hype, analysts and IT leaders said features aren't delivering equal returns in all organizations. Instead of pursuing them all at once, they said CIOs should prioritize which to pursue now and which to adopt later based on organizational needs, IT department objectives and their levels of data maturity.</p> <section class="section main-article-chapter" data-menu-title="Top 5 ITSM trends in 2026"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Top 5 ITSM trends in 2026</h2> <p>There are multiple ITSM trends worth watching, experts said.</p> <p>These trends have arisen in response to CIO needs, specifically rising complexity, ticket volume, pressure to lower costs, <a href="https://www.techtarget.com/searchunifiedcommunications/answer/Hot-desking-vs-hoteling-Whats-the-difference">hybrid work environment</a>, faster release cycles, demands for better employee experiences and rising expectations to use AI to gain efficiencies, said Sue Bergamo, a longtime IT and cybersecurity leader who served as a trustee with the Boston chapter of the Society for Information Management.</p> <p>Here's a closer look at five notable ITSM trends.</p> <h3>1. AI in ITSM</h3> <p>The availability of AI capabilities in ITSM is nearly universal, but their use by IT departments hasn't been as widespread to date. That's now changing.</p> <p>"Adoption is growing quickly, though most organizations are still early‑to‑mid maturity," said Bergamo, who provides fractional CIO and chief information security officer services through BTE Partners and now serves as general partner and CISO at Cyber Scale.</p> <p>Figures from a May 2026 <a target="_blank" href="https://www.teamdynamix.com/download-ai-itsm-market-study/#get_marketstudy_cta" rel="noopener">report</a>, "The State of AI in ITSM" from ITSM provider TeamDynamix, confirm as much.</p> <p>The report, which was based on responses from 392 IT professionals, found that 87% of organizations use or plan to use AI in ITSM within the next two years. More specifically, the research showed that 55% were piloting AI, 28% were in the deployment phase and 4% were optimizing it.</p> <p>Among those who are using AI, 88% use it AI for knowledge management (KM), gap identification and content creation; 82% use it to power virtual agents to deflect tickets; and 71% use AI to accelerate resolution through improved triage.</p> <p>Bergamo said she also sees IT using AI to improve ticketing systems with automated routing, categorization, virtual agents, summaries and <a href="https://www.techtarget.com/searchcontentmanagement/definition/workflow-automation">workflow automation</a>.</p> <p>Additionally, the TeamDynamix research found that the IT departments using AI are seeing benefits, with 82% reporting ticket deflection, 71% reporting reduced resolution times and 76% reporting improved customer satisfaction.</p> <p>However, Bryce Austin, who offers CIO, CISO and consulting services through his firm TCE Strategy, said IT organizations must have the data readiness to make effective use of AI.</p> <p>"If you give the AI engine enough information to go on, it can be quite good," Austin said. But it "can also be a waste of time" for IT departments that lack the data and AI readiness to make effective use of the AI components offered by ITSM platform makers.</p> <h3>2. The integration of ITSM and FinOps</h3> <p>IT shops use ITSM to ensure they're deploying, operating and managing technology and digital services optimally.</p> <p>CIOs are tasked with doing all that cost-effectively. So, CIOs are increasingly adopting <a href="https://www.techtarget.com/whatis/definition/FinOps">FinOps</a>, a framework to manage technology costs -- including those from cloud computing and AI use.</p> <p>ITSM now incorporates that, said Jeff Orr, director of research for IT and technologies and head of the CIO and IT leader research practice at ISG Research. "Financial management is an ITSM pillar, which is increasingly becoming FinOps," he added.</p> <p>According to The FinOps Foundation's 2026 "State of FinOps" report, collaboration between FinOps and ITSM teams <a target="_blank" href="https://data.finops.org/" rel="noopener">has more than doubled</a> since 2023 -- going from 24% in 2023 to 56% in 2025 -- and the percentage of integrated teams has increased fourfold, from 3% in 2023 to 12% in 2025.</p> <h3>3. ITSM as an operational layer</h3> <p>ITSM is essentially a system of record that captures IT's institutional knowledge and details of IT activities. As Orr noted, ITSM encompasses all the guidelines, policies and processes that enable IT -- and therefore, the business -- to operate.</p> <p>Of course, ITSM platforms also support and enable IT actions and activities around those guidelines, policies and processes.</p> <p>But ITSM platforms are evolving to do even more.</p> <p>Using automation enabled by machine learning and AI, ITSM platforms are now coordinating across different systems, Orr said.</p> <p>"It's that change which is creating an operational layer," he said, explaining that an ITSM platform can look at various aspects -- from service delivery to policies, regulations and governance -- and combine them with visibility across different clouds and infrastructures to keep the environment safe, secure, optimized and cost-efficient.</p> <p>"AI technology and automation are providing insights to the organization, providing answers to the question, 'Did you know this is going on in your organization?'" he said.</p> <p>"ITSM is becoming a system of action," according to Orr, who has also called this "the end of passive IT service management."</p> <h3>4. Evolving ITSM to enterprise service management</h3> <p>True to its name, ITSM is all about IT services: the design, delivery and improvement thereof.</p> <p>But as IT and digital environments have become <a href="https://www.techtarget.com/searchitoperations/news/366619261/IT-orgs-expand-business-observability-beyond-their-own-walls">intertwined with business functions</a> in inseparable, indistinguishable ways, ITSM is evolving into enterprise service management.</p> <p>ESM platforms offer a unified portal, providing a single point where employees can request services that require action across multiple departments, not just IT -- such as expanded access credentials tied to a promotion.</p> <p>ESM platforms also offer cross-department workflow automation, universal KM, enterprise-wide incident management, business service analytics and other capabilities that extend components from IT to the overall enterprise.</p> <p>"Enterprise service management is ITSM for non-IT departments," Orr said. He noted that ESM platforms can support the unique requirements of various functions, such as the specific data privacy requirements associated with personnel data in HR, while breaking down the siloed workflows that have persisted even for cross-functional service requests.</p> <p>Orr described ESM as a subset of ITSM, explaining that ESM includes ITSM, but not all ITSM platforms include ESM. He said organizations approaching the end of life for their existing ITSM platforms face the choice between moving to ESM and staying with vendors that provide only ITSM.</p> <p>Worldwide Market Reports estimated the global ESM platform <a target="_blank" href="https://www.openpr.com/news/4296856/future-outlook-of-enterprise-service-management-esm" rel="noopener">market size at $6.2 billion</a> in 2025 and predicted it will reach $12.1 billion by 2032, for a CAGR of 10.1% from 2025 to 2032. It attributed that anticipated growth to "organizations increasingly [recognizing] the importance of streamlining processes across various service domains."</p> <h3>5. XLAs, not SLAs</h3> <p>Although UX isn't one of ITSM's traditional pillars, it has long been important to IT's success. The shift from service-level agreements (SLAs) to experience-level agreements (XLAs) reflects the growing importance of UX in IT service management.</p> <p><a href="https://www.techtarget.com/searchcio/tip/SLA-vs-XLA-Which-one-measures-what-matters">Swapping XLAs for SLAs</a> has been happening for a few years, and it's becoming more prevalent in 2026, with Bergamo explaining that it moves IT to "outcome-based ITSM," which "focuses on the experience and results IT delivers -- not just whether a ticket met an SLA."</p> <p>Research shows that shifting to XLAs helps drive user productivity and satisfaction, as well as continuous improvement within IT delivery, better than the use of traditional SLAs.</p> </section> <section class="section main-article-chapter" data-menu-title="Prioritizing adoption of ITSM trends"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Prioritizing adoption of ITSM trends</h2> <p>These and other advancements in ITSM are having positive effects on IT departments and the organizations they service, experts said.</p> <p>Bergamo said the benefits include the following: &nbsp;</p> <ul class="default-list"> <li>Faster resolution time.</li> <li>Reduced workloads.</li> <li>Better cross‑team collaboration and communications.</li> <li>More automation and fewer manual workflows for greater speed and efficiency.</li> <li>More visibility into business impact.</li> </ul> <p>For IT organizations looking to invest, experts said all the 2026 trends can yield benefits; however, they noted IT teams must have the change management, resources and maturity -- especially with data -- to ensure successful adoption of more AI capabilities, FinOps for AI and even XLAs instead of SLAs.</p> <p>In terms of a priority list, Bergamo listed AI‑driven automation, which provides immediate efficiency gains and reduces operational load; DevOps/<a href="https://www.techtarget.com/searchsecurity/tip/Shift-left-with-these-DevSecOps-best-practices">DevSecOps integration</a>, which also improves speed, stability and cross‑team collaboration; and outcome‑based ITSM, or XLAs, as more immediate to-dos for their ability to improve productivity and experience.</p> <p>Low‑code/cloud‑native modernization and ESM ranked No. 4 and No. 5 on her priority list.</p> <p>However, she and other experts said CIOs should prioritize which new ITSM capabilities to adopt first based on their organization's needs, goals and size.</p> <p>They said that what's more important than deciding which to adopt first is ensuring they adopt new capabilities to evolve their ITSM. That evolution, experts stressed, is key to continuous improvement and to transforming how IT operates.</p> <p>"That then leads to better decision-making about where your IT funds should be going to make IT delivery better, more efficient and more secure," Austin added.</p> <p><i>Mary K. Pratt is an award-winning freelance journalist with a focus on covering enterprise IT and cybersecurity management.</i></p> </section> ITSM in 2026 is shifting toward practical AI, FinOps, ESM and XLAs, with experts urging CIOs to prioritize investments based on data maturity and business goals. https://cdn.ttgtmedia.com/rms/onlineimages/money_g1021600178.jpg https://www.techtarget.com/it-infrastructure/feature/ITSM-trends-in-2026-Where-to-invest-vs-where-to-wait Mon, 03 Aug 2026 14:30:00 GMT <p>As AI transforms the data center industry, its growing resource demands pose significant challenges. The surge in AI capabilities drives higher power consumption and cooling needs, underscoring the urgency of adopting sustainable energy.</p> <p>This article examines the balance between harnessing AI's potential and managing its impact on vital resources while highlighting current challenges and innovative strategies to address them.</p> <section class="section main-article-chapter" data-menu-title="Concerns of increased energy consumption"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Concerns of increased energy consumption</h2> <p>GPUs are the heart of AI technology. GPUs, with their thousands of small cores, are designed to handle massive parallel processing workloads, making them ideal for large-scale AI training. However, GPUs consume 200 watts to 400 watts each.</p> <p>CPUs have fewer cores, run sequential processing and draw power in the range of 65 watts to 250 watts. GPUs are technically <a href="https://www.techtarget.com/searchdatacenter/tip/How-do-CPU-GPU-and-DPU-differ-from-one-another">more efficient than CPUs</a> when measured in floating-point operations per second (FLOPS). GPUs are reported to achieve 50 FLOPS/Watt in some studies, compared with 10 FLOPS/Watt for CPUs.</p> <p>The enormous processing demands of AI require large numbers of GPUs, resulting in very high total power demand during AI "training" mode. When training reverts to low-power "communications" mode, power drops abruptly. Power spikes back up when training resumes, and the cycle continues.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The trend is for AI to push these numbers even higher, with predictions of 1 MW racks and Terawatt-scale data centers, which are beyond the generation and transmission capabilities of almost all localities. </figure> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>The impact on our power infrastructure is huge. A few years ago, power levels of 80 kW to 150 kW per rack were considered eye-opening, and total data center loads of 10 MW to 50 MW were considered high-end. Today, we see 250 kW to 500 kW, half a Megawatt, per rack and total data center loads in the Gigawatts, or thousands of Megawatts.</p> <p>The trend is for AI to push these numbers even higher, with predictions of 1 MW racks and Terawatt-scale data centers, which are beyond the generation and transmission capabilities of almost all localities. Power "swings" can reach 10 MW in milliseconds.</p> </section> <section class="section main-article-chapter" data-menu-title="Concerns of water scarcity"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Concerns of water scarcity</h2> <p>GPUs are liquid-cooled, but their heat must be transported outside the facility. Water has historically been the medium for this because it's several thousand times more efficient than air at removing heat. But that efficiency comes at the expense of evaporation, and huge data centers can vaporize thousands of gallons of water per hour. That's made water usage the most challenging issue the industry is tackling today. There are ways to generate more power, but <a href="https://www.datacenterknowledge.com/data-center-site-selection/virginia-report-finds-groundwater-running-dry-for-new-data-centers">water is a finite resource</a>, and, in many climates, it is simply scarce. Unfortunately, some of those climates are otherwise great locations for large data centers.</p> <p>The Green Grid developed a metric called Water Usage Efficiency (<a href="https://www.datacenterknowledge.com/cooling/a-guide-to-data-center-water-usage-effectiveness-wue-and-best-practices">WUE</a>). It's similar to the well-known <a href="https://www.techtarget.com/searchdatacenter/definition/power-usage-effectiveness-PUE">PUE</a> metric, but, like PUE, it applies only to operating facilities. It does not enable authorities to set or enforce usage standards, particularly in advance of construction, so addressing water usage has become a priority for every Standards and Code committee.</p> <p>Recognizing that water demand can lead to project denial, most large new data centers have reverted to air cooling. This solves the water-use problem but at the expense of significantly higher power demand. Newer magnetic-bearing air-cooled chillers deliver greater efficiency, but the overall power demand without water cooling remains substantially higher.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Concerns over increased noise pollution</h3> <p>A significant amount of effort is focused on addressing the power and cooling requirements of AI, but another important concern is <a href="https://www.techtarget.com/searchdatacenter/tip/Understanding-the-impact-of-data-center-noise-pollution">noise pollution</a>. The numerous air-cooled units used in these facilities can generate substantial noise, hindering acceptance by local municipalities.</p> <p>Most jurisdictions have noise regulations in place, but complying with these limits can be challenging. For instance, sound barriers designed to reduce noise may obstruct airflow, thereby reducing cooling efficiency. The support and infrastructure for AI development still have a long way to go, especially in a rapidly changing environment.</p> </div> </div> </section> <section class="section main-article-chapter" data-menu-title="Offset grid strain with power plants"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Offset grid strain with power plants</h2> <p>The most efficient approach is for facilities to bring their own power. Large new facilities are either constructing their own power plants or buying and reactivating old ones. The <a href="https://www.utilitydive.com/news/constellation-three-mile-island-crane-nuclear-ferc-waiver/821836/">Three Mile Island nuclear power plant</a>, set to open in 2027 to meet the demands of Microsoft data centers and the grid, is an example.</p> <p>Still, most decommissioned plants are, unfortunately, coal- or gas-fired. That increases greenhouse gas emissions. In the best of circumstances, new data centers will build <a href="https://www.techtarget.com/searchdatacenter/tip/How-to-use-data-center-wind-turbines-for-sustainable-energy">wind</a> or <a href="https://www.techtarget.com/searchdatacenter/tip/Solar-shift-How-data-centers-can-embrace-renewable-energy">solar</a> plants, but these have variable capacity and only supplement local generation. Building fuel cells at waste dumps that emit sufficient methane would be ideal, but it requires very careful site selection.</p> </section> <section class="section main-article-chapter" data-menu-title="Managing DCIM software and power spikes"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Managing DCIM software and power spikes</h2> <p>Power spikes are another challenge. Whereas conventional data center power demand changes only a few percent over seconds or minutes, the instantaneous swings in AI facilities are enormous. Power infrastructures are not designed for this. Consider the instability at home when power is restored to a neighborhood after a regional outage. Now imagine this happening and destabilizing the entire power grid many times a day. AI data centers require special power designs to negate these effects.</p> <p>AI is also being used, in essence, to manage itself. <a href="https://www.techtarget.com/searchdatacenter/feature/A-close-look-at-DCIM-software-and-the-broad-vendor-options">DCIM software packages</a>, now a necessary tool for operators to keep control of large, complex facilities, all incorporate some form of AI.</p> <p>Managing and optimizing power and cooling in real time is one feature, with AI used to anticipate transient loads. That helps, but a more complete solution would add bidirectional UPS. Using newer active front-end rectifiers and fuzzy logic or model predictive control, these systems not only maintain reliable power to the servers but also maintain a constant battery charge, allowing them to return power to the grid when voltage drops. This stabilizes the grid and can also reduce energy costs by peak shaving and <a href="https://www.techtarget.com/searchdatacenter/tip/How-data-centers-can-help-balance-the-electrical-grid">assisting the grid</a> during peak demand. Due to the large number of discharge/recharge cycles, lithium-ion batteries are proving to be the battery of choice in these systems. However, newer technologies, such as ultracapacitors, are being explored.</p> <p><em>Robert McFarlane is senior principal in charge of data center design for the international consulting firm Shen Milsom and Wilke LLC. McFarlane has spent more than 40 years in communications consulting and has experience in every segment of the data center industry.</em></p> </section> Discover how AI's demands are affecting data centers' power and water resources, cooling challenges and the approaches being implemented to address these concerns. https://cdn.ttgtmedia.com/rms/onlineimages/disasater_recovery_g1189823155.jpg https://www.techtarget.com/it-infrastructure/tip/Understanding-AIs-resource-needs-and-sustainability-challenges Mon, 03 Aug 2026 14:10:00 GMT <p>Modern networking is transitioning from its hardware-centric roots to a software-driven infrastructure as enterprises race to achieve digital transformation goals and objectives.</p> <p>Managing hybrid, multi-cloud and distributed environments using traditional manual methods has become a growing challenge. For IT leaders, using a network-as-software approach offers significant business value, ranging from greater operational agility to more efficient infrastructure management.</p> <p>This article examines why network management is shifting from hardware-driven to software-driven workflows. The result is a modern operating model that enables automated execution and comprehensive network service delivery.</p> <section class="section main-article-chapter" data-menu-title="From hardware administration to programmable infrastructure"> <h2 class="section-title"><i class="icon" data-icon="1"></i>From hardware administration to programmable infrastructure</h2> <p>Enterprise networking is evolving from manual, device-by-device administration to programmable infrastructure managed through software. The technologies driving this evolution include the following:</p> <ul class="default-list"> <li>Infrastructure as code.</li> <li>Software-defined networking.</li> <li>APIs.</li> <li>GitOps.</li> <li>A <a href="https://www.techtarget.com/searchnetworking/tip/How-NetOps-and-DevOps-concepts-tackle-network-problems">consolidation of NetOps and DevOps</a>.</li> <li>Continuous integration/continuous deployment (CI/CD).</li> <li>Network as a service (NaaS).</li> </ul> <p>These technologies let organizations standardize deployments, automate configuration and set the stage for consistent policy enforcement across hybrid and cloud environments.</p> <p>By reducing manual intervention and configuration drift, organizations can improve operational efficiency, scalability and reliability as they lay the groundwork to use software development practices to enhance network services.</p> </section> <section class="section main-article-chapter" data-menu-title="NetOps, DevOps and GitOps reshape network operations"> <h2 class="section-title"><i class="icon" data-icon="1"></i>NetOps, DevOps and GitOps reshape network operations</h2> <p>Network professionals are adapting software development practices to streamline network services and ensure resilience. These practices include NetOps, DevOps and GitOps. Version-controlled infrastructure repositories and <a href="https://www.techtarget.com/searchnetworking/tip/12-network-automation-ideas-to-incorporate-in-your-network">automated workflows</a> improve governance and operational consistency.</p> <p>These approaches result in fewer human errors, easier provisioning, improved visibility and faster rollback of configuration changes. They enable scalable, predictable operations to meet today's fast-changing business requirements. NetOps, DevOps and GitOps are the foundations of continuous improvement and effective change management.</p> </section> <section class="section main-article-chapter" data-menu-title="CI/CD and the automation of network change management"> <h2 class="section-title"><i class="icon" data-icon="1"></i>CI/CD and the automation of network change management</h2> <p>Network teams are applying <a href="https://www.techtarget.com/searchsoftwarequality/CI-CD-pipelines-explained-Everything-you-need-to-know">CI/CD principles</a> to enterprise networking to automate infrastructure changes. Automated pipelines verify network changes before deployment by testing configurations against predefined policies, validating dependencies and identifying potential conflicts.</p> <p>CI/CD enables the following:</p> <ul class="default-list"> <li>Policy enforcement that ensures new configurations satisfy security, <a href="https://www.techtarget.com/searchcloudcomputing/tip/A-data-sovereignty-primer-for-cloud-admins">compliance</a> and operational standards.</li> <li>Automated deployment pipelines that apply approved changes consistently across environments.</li> </ul> <p>Together, these capabilities <a href="https://medium.com/@kalpana.hellomail/how-ai-tools-are-quietly-replacing-manual-tasks-at-work-8b77f321dff4">reduce manual review</a>, improve change accuracy and create a more predictable process for managing network infrastructure.</p> <p>For IT leaders, CI/CD provides controlled, repeatable processes for overseeing network change while reducing operational risk and supporting business continuity. They pave the way for a software-driven network services model.</p> </section> <section class="section main-article-chapter" data-menu-title="NaaS as a business model for software-driven networking"> <h2 class="section-title"><i class="icon" data-icon="1"></i>NaaS as a business model for software-driven networking</h2> <p>NaaS lets organizations subscribe to network capabilities rather than rely on traditional hardware ownership. NaaS encompasses programmable infrastructure, software-defined networking and automation technologies. The model supports the consumption-based networking often found in hybrid and multi-cloud environments.</p> <p>NaaS is a natural extension of software-driven networking rather than a standalone technology. It provides organizations with scalability, operational flexibility and simplified infrastructure management, offering them a more dynamic approach to network management.</p> </section> <section class="section main-article-chapter" data-menu-title="How these approaches relate to each other"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How these approaches relate to each other</h2> <p>Each approach enables a different part of their digital transformation, including the following:</p> <ul class="default-list"> <li>Programmable infrastructure</li> <li>A new operating model.</li> <li>Automated execution.</li> <li>A modern network service delivery model that supports business objectives.</li> </ul> <p>Each component plays a role in a network-as-software strategy.</p> <ul class="default-list"> <li><b>NetOps, DevOps and GitOps. </b>How teams manage infrastructure.</li> <li><b>CI/CD and automation.</b> How teams safely and consistently deliver network changes.</li> <li><b>NaaS.</b> How the business consumes resulting network capabilities.</li> </ul> <p>Together, these components transform the organization from a hardware-centric infrastructure to a software-driven operating model.</p> </section> <section class="section main-article-chapter" data-menu-title="Strategic considerations for network infrastructure leaders"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Strategic considerations for network infrastructure leaders</h2> <p>Align network modernization with broader business objectives and digital transformation strategies. This is a shift from technology implementation to executive decision-making.</p> <p>A shift to software development principles requires organizations to address <a href="https://www.techtarget.com/searchitoperations/feature/How-AI-in-training-and-development-can-bridge-IT-talent-gaps">workforce readiness</a>, including <a href="https://www.techtarget.com/searchITOperations/feature/IT-skills-development-strategies-to-close-gaps-in-IT-ops">upskilling network teams</a> and fostering closer collaboration between networking, cloud and software engineering functions.</p> <p>Establish strong governance practices that support policy consistency, change control, security and compliance within programmable environments. These practices ensure deliberate decision-making that aligns with ever-changing business requirements.</p> <p>IT leaders must evaluate modernization initiatives based on measurable business outcomes such as operational efficiency, deployment speed, resilience and reduced operational risk.</p> </section> <section class="section main-article-chapter" data-menu-title="Transition to a programmable networking model"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Transition to a programmable networking model</h2> <p>Migrating to a network-as-software approach isn't a one-time, standalone project; it's a long-term operational transformation. Treat it as an <a href="https://www.techtarget.com/searchnetworking/tip/How-network-modernization-creates-measurable-ROI">ongoing commitment to network modernization.</a></p> <p>Use a phased modernization strategy that prioritizes high-value use cases while integrating automation with existing infrastructure. Incremental adoption helps enterprises manage risk, maximize investment value and prepare for software-defined network systems.</p> <p>Organizations that invest in software-driven, programmable networking today will be better positioned to deliver the agility, resilience and scalability needed to support future innovation and digital transformation.</p> <p><i>Damon Garn owns Cogspinner Coaction and provides freelance IT writing and editing services. He has written multiple CompTIA study guides, including the Linux+, Cloud Essentials+ and Server+ guides, and contributes extensively to Informa TechTarget, The New Stack and CompTIA Blogs.</i></p> </section> Manual methods no longer cut it as enterprises adopt advanced architectural models. Instead, software-based approaches like NaaS and infrastructure as code are gaining traction. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1289411982.jpg https://www.techtarget.com/it-infrastructure/tip/Understanding-the-network-as-software-transformation Fri, 31 Jul 2026 14:30:00 GMT <p>One thing that Kubernetes doesn't do -- at least not natively -- is manage costs. Kubernetes doesn't ensure that the resources available within a cluster are used cost-effectively. Nor does it alert admins to situations where costs may be higher than necessary. If businesses don't monitor costs, they may end up spending much more on deploying containerized applications than necessary.</p> <p>In a world where 82% of container users now run Kubernetes in production as of 2025, according to the <a href="https://www.cncf.io/reports/the-cncf-annual-cloud-native-survey/">CNCF Annual Cloud Native Survey</a> released in January 2026, Kubernetes cost management has become a critical component of broader FinOps strategies. Inefficient use of the CPU, memory and storage resources available within Kubernetes clusters can significantly bloat organizations' IT spending.</p> <p>Read on for guidance as we dive into what makes Kubernetes cost management so challenging and which actionable strategies business leaders can adopt to keep spending in check.</p> <section class="section main-article-chapter" data-menu-title="The complexity of Kubernetes cost optimization"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The complexity of Kubernetes cost optimization</h2> <p>Kubernetes cost management can be highly challenging. This is largely due to the sheer complexity of Kubernetes, which requires organizations to adopt unique cost-optimization tactics that extend beyond classical <a href="https://www.techtarget.com/searchcloudcomputing/tip/Apply-these-FinOps-best-practices-to-optimize-cloud-costs">FinOps strategies</a>. Key complications in Kubernetes cost optimization include the following:</p> <h3>Highly dynamic environments</h3> <p>Most Kubernetes clusters are dynamic because workloads are constantly moving around. Part of what makes Kubernetes so powerful is its ability to migrate applications from one server to another based on factors such as changes in resource utilization patterns.</p> <p>However, Kubernetes's dynamic nature also makes it hard to consistently track costs. To track Kubernetes spending on a per-workload basis, you need to know which servers host each workload and how many resources each workload consumes. Since this varies as workloads migrate between servers and as their resource consumption scales up and down, collecting the data necessary to monitor spending in real time is tough.</p> <h3>Varying Kubernetes deployment models</h3> <p>There are multiple ways to deploy Kubernetes. The three main options include:</p> <ul type="disc" class="default-list"> <li>Deploying Kubernetes on your own infrastructure using an on-prem or private cloud deployment model.</li> <li>Running Kubernetes in a public cloud by setting up a Kubernetes cluster on cloud-based VMs that you manage yourself.</li> <li>Using a <a href="https://www.techtarget.com/searchitoperations/feature/Managed-Kubernetes-services-to-consider">fully managed Kubernetes service</a> -- like Amazon EKS or Azure AKS -- which provides both the host infrastructure and Kubernetes software through a single, integrated offering.</li> </ul> <p>Each of these deployment architectures comes with different cost variables. For example, with an on-prem deployment, a primary cost consideration is the capital investment required to purchase the hardware that hosts Kubernetes. In contrast, with a cloud-based deployment, a business would typically pay monthly for the infrastructure it uses every month to operate its cluster.</p> <p>Due to the variability in pricing models, cost tracking and management strategies that work for one type of Kubernetes deployment may not work for another.</p> <h3>The sharing of resources across projects and teams</h3> <p>A common FinOps goal is the ability to align spending with different teams, users or projects. This provides much more granular insight into costs. With Kubernetes, however, granular cost tracking can be tough because multiple teams or projects frequently share a single cluster. Per-workload or per-container cost monitoring is possible, but it requires the ability to track costs as workloads move across different servers within a cluster and as they scale up and down. This adds significant complexity to the cost management process.</p> <h3>Lack of native cost monitoring features</h3> <p>Kubernetes doesn't provide any built-in cost-tracking or management functionality. It can report data, such as how much CPU or memory a workload is using. It's possible to extrapolate from this data to estimate the workload's cost, but Kubernetes doesn't do that translation natively.</p> </section> <section class="section main-article-chapter" data-menu-title="The gap between FinOps and Kubernetes"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The gap between FinOps and Kubernetes</h2> <p>Kubernetes has been an outlier within businesses' FinOps strategies. In theory, FinOps can help organizations optimize the value of virtually any type of IT platform or workload. But in practice, it has typically been easier to manage and optimize spending for other types of resources, such as cloud-based VMs and databases.</p> <p>The problem isn't just the inherently complex nature of Kubernetes cost management. It's also that FinOps tooling for Kubernetes is more limited than FinOps offerings that manage other types of platforms or workloads. For example, while the major public clouds offer built-in cost management tools -- such as<a href="https://www.techtarget.com/searchcloudcomputing/tip/Top-AWS-tools-for-cloud-cost-forecasting"> AWS Cost Explorer</a> -- that track spending for most types of cloud services, they usually report only very basic data about Kubernetes.</p> <h3>Common examples of Kubernetes FinOps failures</h3> <p>To contextualize how organizations can inadvertently waste money on Kubernetes, consider the following common examples of cost-inefficient deployments.</p> <ul type="disc" class="default-list"> <li><b>Too many nodes.</b> A Kubernetes cluster that includes more nodes than are necessary to support its workloads. The organization has to pay for each node, even though some would likely sit idle or be under-utilized in this scenario.</li> <li><b>Excess reservations.</b> A set of containers that have more CPU or memory reserved for them than they need. The business pays for more resources than its workloads use.</li> <li><b>Incorrect instance type.</b> A cluster that uses standard, pay-as-you-go cloud server instances instead of taking advantage of more cost-effective reserved instances.</li> <li><b>Replica issues.</b> A configuration where admins deploy more application replicas than necessary to achieve a workload's reliability requirements. Replicas are a Kubernetes feature that runs multiple copies of the same app to ensure availability but can drain resources.</li> <li><b>Unused persistent volumes.</b> The existence of persistent volumes (PVs) that are no longer actively being used. PVs are a way of storing data persistently in Kubernetes. Typically, admins should delete them when they're no longer in use to avoid wasting storage resources.</li> </ul> </section> <section class="section main-article-chapter" data-menu-title="Best practices for optimizing Kubernetes spending"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Best practices for optimizing Kubernetes spending</h2> <p>The good news for businesses aiming to maximize the ROI of Kubernetes is that doing so is possible, with help from practices and tools designed for the purpose. Here's a look at actionable best practices for maximizing the ROI of Kubernetes.</p> <h3>Reduce server costs</h3> <p>One of the simplest and most impactful steps a business can take to cut its Kubernetes infrastructure costs is to use lower-cost servers. In a cloud-based Kubernetes cluster, reserved instances offer discounted pricing in exchange for a commitment to long-term use. Since Kubernetes clusters are typically operational for an indefinite period, it makes sense to use reserved instances.</p> <p>In some cases, businesses could even use spot instances, which provide very steep discounts but can shut down without warning. While you wouldn't want to use a spot instance for a workload that requires high availability, it can help reduce Kubernetes costs for workloads that can be interrupted, such as application testing or AI model training.</p> <p>Reducing infrastructure costs in an on-prem cluster can be more challenging because businesses can't simply "swap and replace" virtual servers, as they would in the cloud. Still, strategies like extending the life of physical servers or increasing the density of VMs on a physical server can help.</p> <h3>Assess resource quotas and limits</h3> <p>As mentioned above, resource quotas assign a specific amount of resources to Kubernetes workloads. Limits restrict the total resources that workloads can use.</p> <p>Quotas and limits are useful features for efficiently dividing up Kubernetes resources and preventing situations where one workload "hogs" resources from others. However, poorly configured quotas and limits can waste resources and money.</p> <p>Businesses need to revisit resource quotas and limits at least periodically. In the best-case scenario, businesses will continuously compare the quotas and limits for Kubernetes workloads with observability data showing what those workloads are actually consuming. An imbalance -- such as a workload's quota that is significantly higher than its actual use -- requires attention.</p> <h3>Delete orphaned resources</h3> <p>A staple best practice in FinOps is to delete orphaned resources. It's especially important in Kubernetes, since the platform doesn't automatically report or delete resources such as orphaned persistent volumes. Businesses need to audit resources to determine which ones have become orphaned. Admins can do this manually using <a href="https://www.techtarget.com/searchitoperations/tip/Kubectl-commands-and-best-practices-for-new-Kubernetes-users">kubectl</a> -- the Kubernetes command-line administration tool. Also, Kubernetes observability and cost management tools can help automate the practice by identifying orphaned resources automatically.</p> <h3>Separate teams and projects by namespace</h3> <p>Namespaces are a Kubernetes feature that creates virtual clusters. A business can more easily apply different types of resource management and cost-tracking policies with separate namespaces for each team or project that uses Kubernetes.</p> <p>For instance, a namespace that hosts experimental workloads could be configured to use less expensive servers, while more costly, higher-performing nodes are reserved for mission-critical workloads in a different namespace.</p> <h3>Enable autoscaling</h3> <p>Autoscaling capabilities automatically modify resource configurations in Kubernetes based on demand changes. There are three main types:</p> <ol type="1" start="1" class="default-list"> <li><b>Horizontal pod autoscaling.</b> Increases or decreases workload replicas based on workload demand.</li> <li><b>Vertical pod autoscaling.</b> Adjusts workload resource allocations based on workload demand.</li> <li><b>Cluster autoscaling.</b> Adds or removes nodes from a cluster. This type is typically only available in cloud-based Kubernetes environments, since it requires the ability to provision new servers automatically.</li> </ol> <p>In most Kubernetes distributions, autoscaling is not turned on by default, but admins can enable it to help strike a healthy balance between resource allocations and usage.</p> <h3>Deploy Kubernetes cost management tools</h3> <p>Although Kubernetes itself offers no built-in cost management features, add-on tools are available that provide these capabilities. The main offering in this vein is Kubecost, a free and open source tool. Kubecost automatically tracks spending metrics and generates alerts when they exceed the thresholds admins set. The tool can also provide recommendations for optimizing Kubernetes spending.</p> <p>Commercial Kubernetes cost management capabilities are also available in some FinOps tools -- although, unlike Kubecost, most of these don't focus specifically on Kubernetes cost optimization.</p> <p><em>Chris Tozzi is a freelance writer, research adviser, and professor of IT and society who has previously worked as a journalist and Linux systems administrator.</em></p> </section> Kubernetes doesn't track costs natively, and that oversight could be costing the business. Learn how to optimize spending before your budget spirals out of control. https://cdn.ttgtmedia.com/visuals/German/article/cloud-container-openstack-2-adobe.jpg https://www.techtarget.com/it-infrastructure/tip/How-to-overcome-Kubernetes-cost-management-struggles Fri, 31 Jul 2026 10:32:00 GMT 60 webmaster@techtarget.com