Explore real-world edge computing examples for network teams Edge computing vs. cloud computing: What's the difference?

14 edge platforms to consider for a hybrid cloud enterprise

Edge platforms can benefit organizations by bringing resources closer to the user, leading to real-time decision-making, improved response times and increased efficiency.

Choosing the best edge computing platform for a particular organization requires first understanding the nuances of what is meant by edge computing.

In the broadest terms, edge computing is a collection of tools that enables faster computing at the periphery of the network, as near to the source as possible, rather than needing to send data back to a central data center for processing.

Edge computing generally refers to platforms that represent tool combinations that provide data processing, analytics, computing and storage where they are needed, such as to a retail location, warehouse or other data generating source. In almost all cases, "the edge" isn't a world of its own but rather the outer perimeter of an information flow that will pass to the cloud and to the data center in a balance that's very user-specific. The "edge" is a point where computing resources can be brought close to the user. Just how close depends on the application.

In some cases, as in situations where computing is linked to industrial processes, the edge will be very close, usually in the same facility, because any delay in processing events and returning responses could interfere with the application. In other cases, especially cases where the users of the application are distributed, the edge could be a point of cloud hosting centrally located to serve that distributed community. But wherever the edge is and however it's hosted, it's almost always a part of that information flow.

Why is edge computing relevant?

Data centers have existed for about 70 years. Distributed applications like retail point-of-sale (POS) to stores have been around for over 30 years. Cloud computing has been in use for perhaps 15 years. So, what drives interest in the edge?

The answer is that some applications require very close synchronization with real-world activity. In nearly all cases, we'd call these internet of things, or IoT, because these applications use special sensors to collect real-world conditions, special devices to control real-world processes, and both need to be tightly coupled to that real-world activity or damage or injury could result. IoT generates a whole new concept, that of the control loop, where event handling takes place, but this control loop is almost always linked to a traditional transaction flow as well.

Edge computing extends cloud and data center computing. Exactly how that's done depends on the application's relationship with the real world and with users.

Almost every user of edge computing and almost every provider of tools to support it link it with IoT. Some of these tools can also help build the distributed computing applications that IT teams are already familiar with, like retail POS, but the IoT requirements are what shapes edge computing's applications, and what defines the platforms best suited to support it.

For those reasons, enterprises consider edge computing to be almost exclusively tied to IoT, because IoT applications involve the processing of events generated by real-world activities, processing that will generate a command that then influences that same activity. It might be opening a gate, starting a machine, or turning on and off lights. How the real-world activities are distributed -- in a facility, a community or across a country -- determines the scope of "the edge" and where it should be hosted for optimum performance and cost. The application defines the edge, and the edge platform hosts the application in the appropriate place.

This report divides edge computing into categories based on whether the applications indicate that the edge is an extension of the cloud, an extension of the data center or an extension of the real-world processes themselves. The ranked lists in each category are based on input from nearly 300 enterprises across 30 verticals and in 12 countries, and on expert software architect views. Rather than identifying every possible edge computing tool, the report addresses the edge platforms enterprises most often cite.

The edge is a part of an information flow that starts with the user, includes enterprise data repositories, and turns around to the user again. Parts of this flow are usually hosted in the cloud or data center already, and so edge requirements are often addressed by extending cloud and data center resources closer to the user. Existing applications and implementations, then, must also be considered in picking the best edge approach. For that reason, this report is divided based on the possible sources of edge technology, from the data center, the cloud and the target real-time processes themselves.

Top edge capabilities from public cloud providers

Most enterprises rely on public cloud hosting for the front-end portion of their applications. Many create hybrids by linking the cloud applications back to their data center for transaction processing, data storage and business analytics using confidential information.

About a third of these companies also deploy computing resources in edge locations, usually at points near the processes being automated, for missions ranging from IoT data collection to retail POS and financial processing. These resources serve as an information bridge between the control of these target processes and the cloud.

If these edge applications connect to the cloud, it's convenient to use the same development teams and tools for both. That means extending cloud platforms to those secondary customer-premises edge sites. For this mission, the best strategy is to link edge development to a company's current cloud applications and cloud providers. Fortunately, all the major cloud providers support this, and this is the form of edge computing that enterprises say is both the largest installed base and the fastest-growing opportunity.

Each of the cloud providers listed below offer both traditional public cloud hosting, and tools to extend that hosting and the cloud services it includes to the premises, near the real-world processes being automated. These are tools that create a local edge from the cloud, and so they're based on the public cloud services they offer.

Two top public cloud providers, Amazon and Microsoft, also top the list of public cloud providers of edge computing. These companies share two critical capabilities: a version of the most prevalent real-time specialized platform for event-driven and IoT systems, and a managed-edge service option for distributed computing applications like retail POS.

Tied for 1: Amazon

Amazon's edge strategy has two options. AWS IoT Greengrass and AWS Outposts provide on-premises hosting of elements of AWS services. AWS Local Zones and AWS Wavelength provide off-premises edge-of-cloud capabilities. Greengrass is the on-premises piece offered in an open source form and Outposts is essentially a managed service form of Greengrass. Amazon also offers a complete IoT suite with more than a dozen tools to facilitate development, deployment and management of both IoT devices and software. Since IoT is the dominant edge application, these tools are designed to facilitate building edge applications either on premises via Greengrass or at the edge of the cloud, via Local Zones and Wavelength.

There is a public-cloud AWS component of Amazon's edge strategy consisting of tools to deploy an on-premises AWS as a managed service through Outposts and deploy edge-of-cloud elements through Local Zones. These tools allow cloud deployment and management practices already used by enterprises to deploy and manage edge applications.

Amazon also offers an appliance called AWS Snowball Edge for edge hosting to facilitate easy deployment. Enterprises, whether they use Amazon's edge services or not, say that Amazon has the richest set of edge capabilities, so those who need the broadest possible set of edge and hybrid features should look at Amazon first.

Tied for 1: Microsoft

Microsoft has an event-centric set of nearly a dozen tools labeled for IoT but more broadly useful in edge applications. These include Windows for IoT, which lets Windows systems run IoT applications. Because this opens the full range of Windows development and deployment tools to IoT or edge developers, it's a primary reason why enterprises like the Microsoft Azure edge features. Azure IoT Edge's service provides the ability to extend Azure services to on-premises edge devices. Azure Stack Edge is an on-premises extension of compute and storage services to the edge. Users rate Microsoft's edge tools as the easiest to use and to integrate. It's also top-rated among new cloud users.

2. Google

Google has a different edge path, abandoning its IoT-centric tools in favor of a broader strategy based on open source container software developed by Google and now supported in the Cloud Native Computing Foundation.

Google Distributed Cloud Edge is a Google-managed, on-premises-hosted option that has the tightest possible integration with Google's cloud services. This makes Google Distributed Cloud Edge a choice for users who see edge computing more as an extension of the cloud than as the hosting point for emerging IoT processes that connect back to traditional transaction processing applications. IoT applications are supported, but most enterprises report using Google's edge in distributed computing missions like retail.

In IoT applications, Distributed Cloud Edge usually pairs with specialized edge hosting tools. Those tools support applications that manage the low-latency control loops and serve as a bridge between IoT control and traditional transactions. Google treats the edge as a managed extension of the cloud, so it's most suitable for users who want to build their IoT and edge applications as cloud extensions and who have the skill to develop applications on specialized tools. Prospective edge users should take special care in selecting Google's strategy because of its highly specialized approach.

3. IBM

IBM has a cloud-centric approach to edge computing, promoting the use of IBM management tools and services to unify the edge with IBM's vision of hybrid and multi-cloud. Their primary goal is to integrate IBM and Red Hat edge products with IBM's cloud services. IBM's own contribution to this is its Edge Application Manager designed to operationalize the deployment and management of edge elements wherever the "edge" is defined.

Enterprises say that IBM's greatest asset at the edge is its vision of the hybrid cloud. Most who have reviewed or adopted IBM's cloud and edge products are IBM customers at the software level, Z series mainframe users or Red Hat platform users of IBM's professional services. For enterprises not already IBM cloud customers, it's best to view IBM's edge offerings in the form of its Enterprise Platforms for Edge Computing tools, in the following section.

4. Oracle

Oracle relies on Roving Edge Infrastructure for its edge strategy. Roving Edge Infrastructure consists of server and software platform tools that extend Oracle Cloud. These tools can both pre-process information locally, enhancing performance and efficiency, and stand in for the cloud in case of an outage. The Oracle strategy is most appealing to enterprises that are already customers of its cloud services. But Oracle's Roving Edge approach is the most generalized of all the cloud provider edge models. It includes processing of content, AI and machine learning, remote and distributed computing, and IoT.

For IoT, Oracle's Internet of Things Cloud Service provides a highly structured and integrated link between IoT devices and Oracle Cloud, suitable for virtually any device type, including phones and Java processes. This tool is the primary point of enterprise interest in Oracle's edge, other than for users of Oracle Cloud. Its versatility means it can link almost any specialized edge technology to Oracle Cloud. This functionality benefits organizations with local edge hosting, perhaps as part of an industrial system, that want to link it to a public cloud service.

5. Salesforce

Most edge computing deployments are justified by the promise they'll improve access to real-time applications to overcome latency constraints. Salesforce is interesting not only because that's not what its edge strategy aims at, but also because it's a form of "virtual edge" that doesn't rely on on-premises hosting.

Salesforce is primarily a SaaS provider and often markets directly to line departments rather than IT organizations. For line departments that need an edge approach that supports better performance for traditional applications rather than newly developing applications, the Salesforce Edge Network is their answer.

The goal of the Salesforce Edge Network is to connect a Salesforce SaaS user to the nearest Salesforce Edge Network hosting point, regardless of the user's location. This approach anticipates cloud-provider edge hosting as Salesforce deploys new close-to-the-edge locations that can be connected via the virtual edge. Some enterprises could use the same approach to integrate local customer-provided edge hosting of Salesforce features if Salesforce decides to offer them in the future. There is currently no commitment to do that, though.

Enterprise platforms for edge computing

There are several vendors offering operating systems, middleware, development tools, and operations and monitoring tools that will run in the data center and the cloud. Organizations often use these same tools to build an edge strategy. The "edge" might be a part of the data center, distributed to remote office locations or linked to manufacturing, warehousing, industrial or utility processes at specialized locations.

The vendors in this section tend to dominate this platform category because enterprises already use them in the data center. In addition, these same vendors offer tools to deploy application components in the public cloud without requiring modifications to reflect differences in the APIs of cloud services. That makes these edge platforms ideal for multi-cloud users.

1. VMware

Now acquired by Broadcom, VMware is well-known for data center virtualization. The company's Tanzu suite is tied with Red Hat OpenShift in enterprise usage. However, VMware is more platform-focused, particularly on virtualization, deployment and operations. Due to VMware's contributions to data center virtualization, the company has gained traction with enterprises in hybrid and multi-cloud applications, where VMware Tanzu tools run on public cloud virtual machines. Both Amazon and Microsoft offer a VMware service targeting users who want to run the same VMware platform in the cloud and on data center equipment, using cloud-provider tools.

The VMware Edge Compute Stack is a complete edge platform suitable for local edge, cloud edge, public cloud, and hybrid cloud and data center. VMware touts its vSphere ESXi hypervisor's real-time virtual machine hosting for edge applications as equivalent to bare metal, and it also has its own version of a real-time operating system (RTOS) for local-edge hosting where latency control is critical. VMware seems to focus the former on telecom applications, but some enterprises use it for latency-critical applications of their own.

Enterprises credit VMware with the most complete, mature and flexible data center-centric edge strategy. The company's edge strategy is an element of its Cross-Cloud Services portfolio and means that it fits automatically into a multi-cloud enterprise.

The recent acquisition of VMware by Broadcom is changing product strategies and channel relationships, in particular. Because of this, prospective edge users considering VMware should review the current state of the technology and the channel relationships through which edge technology is supplied, before making a final decision.

2. IBM Red Hat

The IBM and Red Hat combination created a top player in the enterprise edge platform space. While IBM isn't a top name in the cloud, it has a respectable cloud business. Its hybrid and multi-cloud emphasis, combined with its loyal customer base, made it a player. The acquisition of Red Hat enhanced that position by creating a link between IBM and the massive base of Linux software and customers.

IBM's edge computing strategy combines its Power S series servers, which support Linux, the company's UNIX variant called AIX, or its integration-centric IBM i operating system. While enterprises currently use Linux or AIX most often, there's considerable interest in the IBM i system, a unique edge platform based on object-oriented programming principles. Unlike most real-time specialized operating systems, IBM i can run on almost any platform, including the public cloud and virtual-machine edge services.

Red Hat builds its edge strategy on a combination of its Enterprise Linux operating system, the standard version or with an optional real-time kernel; its OpenShift development and operations platform; and Validated Patterns, which are recipes for application models that are assembled and integrated by Red Hat and available for easy adoption by enterprises. While these will run on IBM Power servers, enterprises use the Red Hat edge platform tools on virtually all the standard servers and all the major public cloud services.

OpenShift is widely used in the data center. Because it's supported on public cloud services as well, OpenShift is widespread among multi-cloud enterprises that want a single software toolkit across all their cloud providers, rather than cloud-specialized service tools. Red Hat's edge strategy is a direct extension of its multi-cloud, hybrid cloud and data center strategy, and this approach is increasingly popular with enterprises. Combine this with IBM's AI and professional services, and you have an exceptionally versatile edge capability.

3. HPE

Hewlett-Packard Enterprise (HPE) is the top edge player, offering a wide range of both software and hardware platforms, as well as deployment and operations tools. Enterprises say that HPE's edge strategy, based on HPE Edgeline hardware and integrated software, is virtually a turnkey approach. Edgeline is a product that could arguably be placed in the "Specialized Edge" category. HP is a company that applies the broadest mission set to the concept of edge computing, including content delivery and distributed computing in applications like retail POS.

HPE has also launched both a hybrid cloud and an edge-to-cloud product, GreenLake, aimed at augmenting its Edgeline real-time edge, mirroring IBM's success with its hybrid cloud positioning, and competing with cloud provider tools that aim at extending the cloud toward on-premises application. GreenLake is a managed edge and cloud platform, and it might gain HPE enough traction to raise its profile in the Cloud Provider Edge category as well, but that's still in the future.

Enterprises find HPE sales and marketing a bit uneven depending on the location and the vertical market involved. But they say HPE has a strong channel program for edge computing applications, and this program is gaining traction with medium-sized businesses that want a one-stop shop for edge computing. Thus, HPE is widespread with SMB edge users and enterprises.

4. Google Kubernetes and Anthos

Kubernetes is the top container orchestration platform for enterprises, and it's a component of many data center and cloud virtualization suites. Anthos is a supplement to Kubernetes, allowing organizations to build multiple Kubernetes clusters and manage them as a single unit.

This approach is attractive to enterprise edge prospects that already use Kubernetes in the data center or cloud and want a familiar and unified strategy for edge application operations management. However, to use these tools at the edge, the edge hosting platforms must be able to run Kubernetes. Many edge hosting applications use bare metal servers and a single software component, not containers, so the Kubernetes/Anthos model is not applicable.

Enterprises that see on-premises edge hosting as an extension to their data center like the Kubernetes/Anthos model because it doesn't change their deployment and tech support practices, and the tools are well-documented with widely available skilled engineers in most labor pools. Any enterprise that uses Kubernetes should assess Kubernetes and Anthos for edge computing needs.

5. Dell

Dell is best known as a hardware provider and well-known in the server market. So it's not a surprise that Dell's edge vision is based on the extension of its data center server strategy to edge locations, making it a kind of distributed data center or distributed computing model. This offering is the value enterprises see in Dell's edge strategy; it doesn't mandate or depend on any major changes to the development and operations strategies they're already familiar with.

Dell's edge approach can be applied even by enterprises that don't currently use Dell products, since open components form the basis of Dell's software platform. Dell can offer a complete edge strategy without requiring integration by providing hardware and software. It also has a line of Edge Gateways for IoT based on Intel Atom processors that can provide integration with existing or new IoT applications based on other vendors' technologies. The open source EdgeX Foundry software platform is the basis for these systems.

While Dell is currently in fifth place, enterprise interest in its approach is growing.

The specialized edge: Embedded control and real-time

One point users of, and even candidates for, edge computing make clear is not all applications that may involve local hosting are IoT applications that require stringent control of process latency. Retail POS applications deploy computing close to operations, but edge purists see this as nothing more than distributed computing, an alternative to using terminals that link all the way back to the data center and are vulnerable to communications outages. They say the "real edge" is where real-time event processing occurs, processing that is local to the activity because it's controlling the activity and must be fully synchronized with it.

That real-time processing is almost always linked to IoT, in the form of event-to-response or "control loop" processing found in manufacturing, transportation and utilities. Applications like this are usually supported by specialized devices, tightly coupled to industrial processes, so the software is sometimes called real-time and sometimes embedded control. Developing this kind of edge application requires special skills and tools. Many edge platforms simply don't have enough users to gain broad support among software providers.

There are two realistic paths users can take to the specialized edge, one based on a real-time-optimized version of Linux and one on the native RTOS. Both have multiple product sources within them, but both have versions enterprises widely favor. RTOS is more customized for low-latency applications and demanding IoT. But the real-time Linux versions have better developer and operations support, and a wider pool of experienced developers. Most enterprises will have in-house Linux capability, so Linux is the better option for distributed computing missions, which are typically more like traditional data center or cloud developments.


A real-time operating system is an operating system customized for real-time, event-driven, highly latency-sensitive applications. This path should be the choice for extreme applications. Overall, there are several dozen versions of RTOSes available. But enterprises looking to develop edge computing applications should focus on the market leaders, where support and experienced developers are more likely to be available. The biggest question is whether to get an RTOS from Amazon or Microsoft, as a partner element in their public cloud and edge strategy, or from a software source.

The smart strategy is to use an RTOS version from the organization's cloud provider, where local edge computing should connect with cloud applications. A tight workflow connection between the RTOS edge and the cloud will be easier using a cloud-provider RTOS. A common source will also provide a good set of development and deployment tools. If there is no "primary" cloud provider, if the local edge will connect with the data center or if multi-cloud deployments rely on edge tools, look for an RTOS version supplied by, or affiliated with, the specific enterprise platform to be used.

Enterprises consider Wind River System's VxWorks as the top player in RTOS, followed by QNX Systems' QNX and Green Hills Software's Integrity, when they expect tight coupling between the local edge and the cloud or data center.

2. Linux

Linux is the most well-known server operating system, so developers who are familiar with it are available in almost any labor market. Linux offers a wide variety of software tools for development, deployment and management, and many will run on real-time versions of Linux. It's easy to use standard Linux systems to develop and manage real-time Linux resources. As a result, real-time versions of Linux are favored by enterprises where latency control isn't the paramount issue.

Wind River is the real-time Linux version most recognized by enterprises. Since Wind River also offers a version of RTOS, the company is an objective source of both leading specialized-edge operating systems. Its edge tools are widely used, well-documented and versatile.

An interesting addition to the real-time Linux inventory is Raspberry Pi OS, a derivative of Debian Linux. Many enterprises have been evaluating the Raspberry Pi family of microcontrollers for edge applications because they're inexpensive and system integrators frequently use them in packaged IoT applications.

Representing a growing need

The fact that there are multiple approaches to edge computing and multiple ways of implementing each approach means that managing workloads are often distributed across hybrid edge, cloud and data center systems, and just deciding where to put application components can be complex. One new product from a startup represents a move to address this complexity. It's reasonable to expect features to evolve in other edge portfolios to address these issues.

Arctos Labs

Given all the places where enterprises might decide the edge could be, it's not surprising that there's a growing set of challenges in determining where to host components of an application. Arctos Labs offers an innovative product that can help enterprises find the optimum balance of hosting costs and quality of service across a full range of hosting options, from local, on-premises edge or cloud edge, through the data center to traditional cloud computing.

Arctos divides a hybrid cloud and edge resource pool into compute zones that represent either zones within the cloud, like Amazon's availability zones, or clusters in a data center or at the edge. Each zone has a set of capabilities and constraints, and companies can map deployments based on those parameters. This approach works similarly to using Kubernetes features in targeting deployments, but with a lower overhead than Kubernetes, the ability to explicitly account for latency differences, and without a requirement for containerization.

Tom Nolle is founder and principal analyst at Andover Intel, a consulting and analysis firm that looks at evolving technologies and applications first from the perspective of the buyer and the buyer's needs. By background, Nolle is a programmer, software architect, and manager of software and network products, and he has provided consulting services and technology analysis for decades.

Next Steps

Software developer goals that are top of mind

Dig Deeper on Digital transformation

Cloud Computing
Mobile Computing
Data Center
and ESG