Bare metal Kubernetes: The 6 things you wish you knew before 2022
by Anton Smith on 14 December 2021
2022 is right around the corner, and it’s not just time to prepare for Christmas, play video games, buy presents, or share anti-Christmas memes. It’s time to start making some predictions for bare metal Kubernetes!
Take a minute and let’s think about it. Developers have advent of code so they’re busy right now. Sysadmins and DevOps can play games like predicting what’s going to happen next year for bare metal Kubernetes. And with that, here are our 6 predictions for 2022.
1 – Kubernetes (K8s) will continue becoming the default API for infrastructure
Tools like kubectl and the Kubernetes API are becoming a new standard for interacting with infrastructure. This is due to the growing understanding and adoption of Kubernetes, particularly in public cloud. This trend will accelerate in 2022 as CI/CD and GitOps continue to dominate. Kubernetes will provide the predominant abstraction layer for infrastructure.
There are other places where Kubernetes is desirable aside from public clouds, which brings us to the next point.
2 – Bare metal provisioning will become a standard building block for multi node clusters at the edge
Kubernetes provides a nice, consolidated API for deployment of applications and containers. This is not only desirable for public cloud, but also for edge deployments. Edge deployments usually mean many more sites. Operations teams still want to use the same interface that they use with public cloud.
One major difference with public cloud is that the physical servers need to be managed. Gone are the days where manually deploying and configuring servers is an acceptable way to do this. Therefore, tools like MAAS will become crucial components for edge Kubernetes deployments. Equally important are standardised integrations between Kubernetes and bare metal, such as the Spectro Cloud Cluster API MAAS provider and Juju-MAAS integration.
3 – Single node clusters for edge will be a thing
Edge sites with a single machine will be increasingly important. There are sites where the application does not warrant hardware redundancy or it is not economically viable to do so. However, it is desirable to retain the Kubernetes API for rolling out applications and managing their life cycles.
Small, nimble, production-grade Kubernetes distributions such as MicroK8s are perfectly tailored to small deployments. This is extremely important in order to offer K8s on single node clusters. MikroK8s will provide the consistent packaging and deployment experience that is needed to support hybrid cloud.
4 – Bare metal Kubernetes will be the default for all new 5G base stations
The Telco industry is always working hard to find more efficient ways to roll out applications. The 5G architecture has disaggregated many functions, allowing them to be containerised and rolled out on COTS (commercially available off-the-shelf) hardware. These are often referred to as CNFs (Cloud Native Functions).
Thanks to the maturation of Kubernetes, this allows the deployment and management of CNFs to be more effective. It also allows standardisation on an infrastructure abstraction. As a result, the preferred method for deploying 5G base stations and associated functions will be via Kubernetes.
MicroK8s and MAAS have an important role to play to provide the K8s infrastructure.
5 – AI/ML or VR/AR workloads will be delivered to the edge on bare metal Kubernetes
Expect to see these exciting applications drive more deployments to the edge. Due to the way these applications use hardware such as GPUs, they will are easier to deploy and offer superior performance on bare metal. AI/ML (Artificial Intelligence/Machine Learning) creates efficiencies by being closer to the data it is processing. VR/AR (Virtual Reality/Augmented Reality) also benefits by being closer to users. In the case of VR/AR, it is because of reduced latency to the end user. For AI/ML, processing data into meaningful information closer to where it is generated reduces central storage requirements and lowers network throughput. Game streaming where the game engine runs on the edge will also benefit thanks to lower latency.
Like other applications, deploying these applications to the edge is easier and operationally more efficient when using Kubernetes, so it should be expected that they will also be deployed as containers on bare metal.
6 – Multi-tenancy at the edge will see more open source Virtual Machine (VM) based solutions
VM-based isolation is important when the strictest security is needed. Hosting tenants on shared hardware is the key driver for this. Open source solutions such as LXD will see increased adoption at the edge as a result. LXD has a rich set of functionality and offers the ability to lower CAPEX for VM management.
Containers and orchestration via Kubernetes will still be important. Containers will co-exist with LXD by being deployed inside VMs.
MAAS will be needed to provide bare metal management when there are multiple machines present.
2022 is shaping up to be a very exciting year for infrastructure and bare metal Kubernetes. In particular at the edge, where MAAS, LXD, and MicroK8s have an extremely important role to play.
The increased adoption of Kubernetes as an infrastructure abstraction, combined with the rise of edge computing will cause these production grade technologies to come into their own.
Bring on 2022, and let’s see how these predictions fly.
Canonical has released an extensive whitepaper for bare metal Kubernetes – going in depth into many of the different aspects involved. Download it here!
his is a very exciting time for HPC, as we are seeing a lot of innovation in the space. Servers are taking on advanced capabilities in terms of compute and networking and we are seeing compatibility increase. Overall cluster deployment and usage is being driven by ease of use. […]
“He’s dead, Jim.” Dr. McCoy DHCP is dead; long live DHCP. Yes, the end-of-life announcement for ISC DHCP means that the ISC will no longer provide official support or updates for the software. Our ever-faithful, omnipresent friend — the familiar dhcpd daemon — is retiring, albeit over a really long walk to that cabin in the […]
We want you to join our Ubuntu circle, and help us document MAAS. More minds, more eyes, more hands make better doc. […]