GigaOm Radar for Kubernetes for Edge Computingv2.0

1. Executive Summary

The escalating growth of data generation at the edge has fueled demand for strategic compute capabilities positioned closer to the source. In response, Kubernetes has emerged as an enticing, standards-based platform for constructing applications geared toward processing data at the edge.

As Kubernetes is the predominant standard for orchestrating containers on a large scale, its adoption at the edge seamlessly extends the orchestration and management capabilities that have made Kubernetes a prevailing choice in cloud and data center environments. The diverse array of use cases and challenges encountered in edge environments necessitates a robust solution like Kubernetes to aid customers in navigating the intricacies.

Today, connected and smart devices span a range of industries, from retail and transportation to healthcare and manufacturing. Whether in remote wind farms or autonomous vehicles, embedded computing devices are generating copious amounts of data, data that emanates from locations far beyond the traditional confines of data centers and cloud environments. Processing such data in centralized locations poses considerable challenges due to intermittent and unreliable connectivity, constrained bandwidth, and high latency. The logical solution is to process data in geographic proximity to its source.

Kubernetes stands out as an ideal platform for building the applications that handle this data. Container-based applications, requiring fewer resources than traditional operating systems, prove well-suited for the constrained resources of rugged, embedded-device form factors. In comparison to static approaches like programmable logic controllers, a general-purpose computing platform, such as Kubernetes, provides greater flexibility. Furthermore, Kubernetes adheres to an operating model familiar to data center operators, offering economies of scope and scale. The same application development methods can be seamlessly applied to cloud-based, traditional data center, and edge-based Kubernetes clusters.

The deployment of Kubernetes for edge computing generally follows one of two common approaches:

• Platform: This approach supports a broad spectrum of hardware devices and existing infrastructure components, allowing significant flexibility and choice in both hardware and deployment location enabled by Kubernetes.
• Appliance: Resembling the approach taken in hyperconverged infrastructure, this method combines a highly opinionated selection of compute, storage, networking, and orchestration with a vendor-selected software stack enabled by Kubernetes.

Some offerings may combine aspects of both approaches. Platform deployments prove beneficial when existing infrastructure is in place or when customers prefer to maintain a high level of control over the approach and placement of the solution. In contrast, appliance deployments are advantageous when customers prefer vendors to assume responsibility for the qualification and support of the hardware and software combination for hyperspecific, well-defined use cases.

As data processing capabilities, analytics, and real-time decision-making continue migrating closer to where data originates, the concept of edge computing has evolved to enable new market opportunities. Within edge computing, there are gradations of “edge” that provide different capabilities and cater to specialized use cases.

• The near edge sits between cloud data centers and the extreme edge, consisting of mini data centers like cell towers, central offices, and campus facilities. The near edge provides compute power, storage, and networking closer to users and devices than the cloud, enabling use cases like content delivery networks, data aggregation from internet of things (IoT) devices, and real-time data analytics requiring very low latency. Kubernetes container orchestration is well-suited for near-edge applications since it allows centralized deployment and management of containerized workloads at this intermediate edge layer.
• Moving closer to the data source, the far edge resides on-premises, very close to endpoint devices and sensors. This includes small ruggedized servers or integrated compute sitting inside devices within retail stores, factory floors, and vehicles. The far edge focuses on extremely low latency use cases like AR/VR, industrial automation and control, and autonomous vehicles. Lightweight Kubernetes distributions are optimized to provide the same Kubernetes container benefits but at the resource-constrained far edge.
• Finally, at the furthest reach is the device edge, consisting of the endpoints themselves—various sensors, gateways, controllers, and microcontrollers. These devices collect and preprocess data and communicate with far-edge servers. Being highly optimized for specific functions, device edge elements contain only necessary compute, memory, storage, and power suited for embedded environments. Software like Podman can deploy containerized logic directly on devices without full Kubernetes orchestration.

This is our second year evaluating the Kubernetes for edge computing space in the context of our Key Criteria and Radar reports. This report builds on our previous analysis and considers how the market has evolved over the last year.

This GigaOm Radar report examines nine of the top Kubernetes for edge computing solutions and compares offerings against the capabilities (table stakes, key features, and emerging features) and nonfunctional requirements (business criteria) outlined in the companion Key Criteria report. Together, these reports provide an overview of the market, identify leading Kubernetes for edge computing offerings, and help decision-makers evaluate these solutions so they can make a more informed investment decision.