• Share this article:

K8s at the Edge – Some Context on the New Kubernetes IoT Working Group

Wednesday, September 26, 2018 - 07:00 by Mike Milinkovich

It’s hard to believe that Google released Kubernetes as an open source project only three years ago. What began as the Borg cluster management platform to provide services like Gmail and YouTube at global scale is now the standard orchestration layer at the center of a massive industry shift to cloud native.

Enterprises are being catapulted into system resource engineering concepts that have been the bedrock of operations at web-scale leaders like Google and Netflix, and the open source stack underneath it all is evolving so fast it’s hard to keep up.  Kubernetes is attracting all sorts of exciting frameworks and services (see projects like Istio and Envoy), and no one could have anticipated it having this degree of momentum.

So what does any of this have to do with IoT?

The Eclipse Foundation is excited to be part of today’s announcement of a new Kubernetes IoT Edge Working Group – in conjunction with CNCF and supported by leading vendors like Bosch, Eurotech, InfluxData, Red Hat, Siemens, Vapor IO, and VMware.  

Eclipse Foundation has a lot on the go with IoT — 2.93 million lines of code (by last count), 42 different member organizations, 36 projects and 280 developers. The pace of innovation in IoT is very fast, and there are still a lot of great unsolved challenges at the IoT Edge.

Kubernetes was born in the datacenter. Its original promise was bridging datacenter and cloud.

But Kubernetes has a ton of potential for handling massive workloads on the edge — providing a common control plane across hybrid cloud and edge environments to simplify management and operations. Chick-fil-A, for example, is running Kubernetes at the edge in every restaurant.

Some of the great unsolved problems at the IoT edge — connectivity, manageability, scalability, reliability, security challenges, how to bring compute resources closer to edge devices for faster data processing and actions — are being solved as one-offs by the enterprises that are jumping into IoT. This new working group sees Kubernetes as having great potential as a foundational technology for extending hybrid environments to the IoT edge, and believes that broader industry collaboration on requirements definition around Kubernetes will accelerate broader adoption of IoT. Once again, it’s the rising tide theory of open source.

As Red Hat’s Dejan Bosanac (lead for the Kubernetes IoT Edge Working Group) says:

“IoT and edge applications have many distributed components that don’t usually sit together within the same datacenter infrastructure. There are messaging challenges, security has to be re-invented for every application and service, and there are integration and data locality issues with sidecar services. These are issues that shouldn’t have to be re-invented every time — they should be open source infrastructure with broad industry support. Red Hat and this working group see Kubernetes and other cloud-native projects in its orbit as having broad potential sitting between gateways, edge nodes and cloud platforms. Much like the LAMP stack was instrumental to the web-applications era, this group is focused on accelerating a Kubernetes stack for running cloud infrastructure and distributed components at the IoT edge.”

Some of the initial targets for the group in how it evolves Kubernetes for IoT edge applications:

  • Supporting Industrial IoT (IIoT) use cases scaling to millions of constrained devices a) connecting directly to Kubernetes-based cloud infrastructure (IP enabled devices), or b) connecting via IoT gateways (for non-IP enabled devices)
  • Via Edge nodes, bringing computing closer to data sources to support processing and acting on data sooner. Anticipated benefits include reduced latency, lower bandwidth, and improved reliability. Some example use cases:
    • Deploying data streaming applications to the Edge nodes in order to reduce traffic and save bandwidth between devices and the central cloud.
    • Deploying a serverless framework for using local functions that can be triggered as a response to certain events (without communication with the cloud)
  • Providing a common control plane across hybrid cloud and edge environments to simplify management and operations

The initial focus of the working group will be to flesh out IoT edge computing use cases and see how Kubernetes can be used (and to what extent). Among some of  the requirements identified so far:

  • For IIoT applications, the Kubernetes ingress layer must scale to millions of connections
  • That same ingestion layer must provide first-class support for IIoT messaging protocols such as MQTT (it is primarily HTTP TLS centric today)
  • Kubernetes must support multi-tenancy for environments where devices and gateways are shared

There have been a lot of industry watchers who have been wondering aloud how high the ceiling is for Kubernetes. Container orchestration seems like such a specific technology – how can this technology reach this type of ubiquity? Well one thing is clear – Kubernetes has jumped the datacenter and is now clearly in play in IoT ($1.2T market by 2022, according to IDC) as well. We are excited to be collaborating with the Kubernetes community to enable next generation of IoT applications.

To learn more, head over to the Kubernetes IoT Edge Working Group community page.

Have specific ideas for how you’d like to see Kubernetes evolve to meet your IoT use cases? Check out the this working document where you can get a sense of the progress so far, and the areas where you could possibly contribute your talents.