Nokia is backing a new Kubenet initiative that wants to inject Kubernetes more deeply into telecommunication network automation by using the now 10-year-old container orchestration platform to work that orchestration magic on network routers and switches.

The recently launched Kubenet project is an effort to tap what has been learned from using Kubernetes to orchestrate traditional cloud-native environments to automate networking. Wim Henderickx, who is head of technology and architecture at Nokia, explained that this entails the transfer of those cloud-native concepts to the networking world.

“Kubernetes started with containers and container automation and orchestration, and in the meantime it does [virtual machines], you can do servers, you can do cloud,” Henderickx said. “So the question that we have is, ‘why not networking?’”

The Kubenet answer to that question is based on taking what has been learned from using Kubernetes to manage pods in a container environment. Those pods contain “resources” that provide the information needed to run applications and services.

“Basically, the way you should make the analogy is that a pod for containers is what a config resource is in the case of networking,” Henderickx said. “So, what we can do right now with that is we can interact with a networking device in a declarative way if there is a change on the configuration. For example, if someone manually would change a configuration, the controller would see it and we basically could do corrective actions.”

This would allow Kubenet to operate as an abstracted level on top of a router or switch where automation and orchestration of those devices can be managed in a vendor-agnostic way.

“There is a set of new constructs that we have added inside of Kubenet to basically show how you could leverage Kubernetes to actually build a layer-3 VPN or layer-two VPN, or let's say if people use VLANs, in a vendor agnostic way that is also in an abstracted way,” Henderickx explained, adding, “not only for a Nokia router but for Cisco, Arista, or Juniper or Sonic or any other [network operating system] that is in play.”

That abstraction has been a key use point for Kubernetes, which has allowed cloud-native architectures to construct services that can be easily ported between cloud environments. This is based on the open-source nature of Kubernetes that has been adopted as the de facto container orchestration platform for these environments.

Henderickx did note that the Kubenet approach did not have to use Kubernetes as a base, but it brings familiarity to the process backed by a large volume of development and adoption. This also ties into the symbiotic nature Kubenet shares with the recently launched Nephio project.

The Nephio project was launched just over 2 years ago as a joint effort between the Linux Foundation and Google. The intent of that project is to provide Kubernetes-based cloud-native intent automation and automation templates to make it easier for telecom operators to deploy and manage multi-vendor cloud infrastructure and network functions across large-scale edge deployments.

Nephio is designed to sit on top of a Kubernetes substrate either directly controlled from an operator or a hyperscaler-based platform like Google Config Connector, Amazon Web Services (AWS) Controllers for Kubernetes and Azure Service Operator.

It basically takes Kubernetes’ cloud-native orchestration expertise to allow operators to roll out and manage new services in their 5G and edge deployments. This will allow for faster onboarding of network functions into a production environment that mimics the DevOps process used by hyperscalers.

Henderickx, who is also co-chair of the Nephio project, noted that using a similar Kubernetes-base for Nephio on the 5G control plane and for the Kubenet focus on networking control makes for an easier overall management picture.

“You can now actually get a uniform construct that allows us to leverage a single automation orchestration system, when multiple components are then used to actually execute a given domain,” Henderickx said. “Our vision is then you can use Kubernetes as a way to orchestrate all of these components or you could actually build a layer on top that actually stitches those things together.”

Henderickx added that the Nephio project has had an intent to work more on networking, “but there is no actual activity going on in that space.”

“Given that the expertise that is in Nephio right now is really on the radio and on the core side, we said, okay, it's better to do that separate rather than included because then you have real a focus on networking and networking people,” Henderickx said.

The Kubenet initiative is initially being focused in highly automated environments like the data center, with a larger push in that arena expected to come from the growing use of artificial intelligence (AI) that will put additional pressure on resources. But there is also a potential for Kubenet to work in campus environments.

“We believe that initially like a data center or a campus or things like that is where the technology first will be used,” Henderickx said. “If you have an application where you do a little bit of networking – look at private 5G for example – you need a switch, you need to put some information on there and managing that from a central site and basically configuring multiples of these sites is definitely something that would be applicable.”

That focus also has the Kubenet project in talks with other vendors interested in being able to orchestrate in these multivendor environments. This includes “traditional ones like Arista, Juniper, Cisco, Dell, [Hewlett Packard Enterprise], and then potentially also like Red Hat … plus other cloud infrastructure people and I would not exclude even a Broadcom-slash-VMware at some point,” Henderickx said.

Henderickx added that if the project does gain momentum it could fit into a broader environment like the Cloud Native Computing Foundation (CNCF) or the Linux Foundation Networking (LFN) group, which are “the communities these days where you actually do cloud-native types of development.”

“That's going to then depend on each of those vendors and what they want to achieve,” Henderickx said. “Right now we are more in the education and the mindshare space of, ‘hey, guys, we thought about this.’”