This content is part of the Conference Coverage: HPE Discover 2024 news and conference guide

OpsRamp embraces automation for GenAI workloads

In this Q&A, Varma Kunaparaju, HPE OpsRamp lead, talks about soon-to-come platform updates, which include GenAI capabilities and building trust for AI-driven automation.

LAS VEGAS -- The OpsRamp platform will plunge into HPE GreenLake's private cloud AI initiatives in the months to come, adding generative AI capabilities to automate decisions and providing visibility into customer AI workloads.

These new initiatives are only possible following HPE's acquisition of OpsRamp last year, according to Varma Kunaparaju, vice president and general manager of hybrid cloud SaaS and OpsRamp at HPE. Kunaparaju co-founded OpsRamp in 2014 and served as its CEO until 2023.

OpsRamp remains available as a standalone SaaS but now comes bundled with many GreenLake purchases, making OpsRamp HPE's defacto operations management service for AIOps. HPE unveiled at Discover 2024 in Las Vegas new generative AI observability features are coming to OpsRamp as well as an integration with the CrowdStrike Falcon platform for attack surface visibility in AI pipelines.

Here, Kunaparaju discussed these generative AI initiatives for OpsRamp, how the service aims to build trust with customers and what makes visibility of AI workloads different compared with what's come before.

Editor's note: The following was edited for length and clarity.

Following the acquisition last year, what's been the focus of the OpsRamp team within HPE?

Varma Kunaparaju, vice president and general manager of hybrid cloud SaaS and OpsRamp, HPEVarma Kunaparaju

Varma Kunaparaju: It's two parallel tracks we've run. One is how we take the OpsRamp technology and bring it to the broader HPE ecosystem of customers and partners. We focused on two segments: … One is complete care for customers with high-level warranty and support. The second is taking OpsRamp and giving it to GreenLake customers. Our bundle [offering] post-acquisition includes OpsRamp in every GreenLake consumption model.

The second portion has to do with our leadership position in cloud-native, full-stack observability [and] our network observability with Aruba, Juniper, Cisco and Arista [support]. We want to become the next generation of observability, building automation and taking our automation journey with LLMs [large language models] and copilots.

[The platform should] not just tell IT that something went wrong but, from [AI] learning, [the platform] should take remedial action.

What makes AI visibility a different IT ops challenge compared with security or existing infrastructure?

Kunaparaju: Visibility is Layer 1 and not where the value is. When you visualize, can you [observe] a full-stack observability on AI where you have container workloads running on top of LLM stacks, AI models, infrastructure and Nvidia? That's the full observability for AI workloads we announced today as part of HPE Private Cloud AI.

Difference No. 1 is instrumentation. In data center [observability], you're going to monitor and manage your GPUs, CPUs, networks and such. In the case of AI stacks, you need to optimize your GPUs and networking. It's an instrumentation difference. With Private Cloud AI, we did that optimization.

The next layer is automation of that optimization, and that's what we will be executing toward. You will see more advances in the coming quarters, such as can you move this workload in such a way that you don't need to run it on a GPU anymore so inferencing can be done [on that GPU]?

As a startup, we wouldn't have the luxury of what HPE provides. OpsRamp is collaborating with Hewlett Packard Labs to create a foundational AI model for operations.
Varma KunaparajuVice president and general manager of hybrid cloud SaaS and OpsRamp, HPE

How are you working to protect customers and increase confidence among IT ops teams to use the automation capabilities you're working toward?

Kunaparaju: Our automation framework is built on top of a powerful workflow studio.

It's a Wysiwyg Studio where you can draw, drag and drop all your notes in the workflow. Some of those notes could be a decision point where [a customer says], 'If it comes to this, make sure the team approves it before you move a workload.'

It's up to the operations people to configure the guardrails and approvals. That studio is a native [feature] of OpsRamp. We use that feature for patch management and configuration management.

How do you see the OpsRamp platform evolving to further automate the more challenging decisions for IT ops teams?

Kunaparaju: As a startup, we wouldn't have the luxury of what HPE provides. OpsRamp is collaborating with Hewlett Packard Labs to create a foundational AI model for operations.

We are optimizing this with metrics logs to train the model toward autonomous operations. The inferencing and RAG [retrieval-augmented generation] on top of it will make automated decisions because [our technology] knows that needs to be done.

Most LLMs out there are built on foundational models that are optimized for text interpretation. In our case, you need to understand the topology and understand the [decision] matrix. Using 95% of the CPU in one case is not the same as another, where it may be perfectly fine.

We want the AI to produce a run script that if plugged into our ops query and automation language can take control of the execution of workflows behind the scenes.

Tim McCarthy is a news writer for TechTarget Editorial covering cloud and data storage.

Dig Deeper on IT systems management and monitoring

Software Quality
App Architecture
Cloud Computing
SearchAWS
TheServerSide.com
Data Center
Close