Next Generation Architecture

Brian Dumont
4 min readJan 21, 2022

Bare Metal Kubernetes at the core

What is next generation architecture?

As a Solution Architect my goal is to create architectures for my customers that are innovative, yet stable and cost effective that will serve their business well. When I’m talking to my customers and presenting options I believe that I must understand where the industry is going such that they can be informed on what others are doing given similar use cases.

I truly believe that organizations that are well-informed have the best chance of implementing solutions that will enhance the business value of their IT infrastructure.

While public cloud dominates the conversation (and it should) I believe there still is a need for on-premise infrastructure thereby creating a hybrid environment.

This post is designed to begin the exploration of what the next generation of on-premise architecture will bring.

In today’s IT community there are three primary default architectures that are deployed:

  • On-premise infrastructure based on virtualization with Kubernetes orchestration running in virtual machines
  • Public cloud infrastructures running virtual machines
  • Public cloud infrastructures running Kubernetes

Or most likely some combination of the above.

Mike Barrett from Red Hat recently published a blog talking about the potential benefits of an alternative architecture that involved bare metal at the core. His article went on to talk about two things that allows this architecture to become a reality:

  1. Linux containerization and the ability to have portable workloads packaged in an open container initiative (OCI) format
  2. Intelligent application services such as big data analytics, high-performance computing, machine learning (ML), deep learning, and artificial intelligence (AI).

I would add a third driver and that is the focus on small footprint Edge computing.

I echo Mike’s comment that no matter how broad Kubernetes is adopted there will always be a need for Virtual Machines. Red Hat saw that and created the KubeVirt open source community to help look at virtual machines differently.

In that light the questions I have been asking myself and others is:

  • What will the on-premise infrastructure of the future look like?
  • Will it continue to be dominated by virtualization or will the push toward containerization move this infrastructure toward Kubernetes.
  • And if Kubernetes is the future, does bare metal make a comeback in the data center?

Use Cases for Bare Metal Kubernetes

With a demand for bare metal solutions from customers with Edge computing and CPU/GPU intensive workloads combined with the technical advances in Linux containerization and an innovative open source community is the time right for using Bare Metal Kubernetes at the core of the on-premise infrastructure

I see use cases that make a lot of sense for running Kubernetes (Red Hat OpenShift) on bare metal and layering KubeVirt (OpenShift Virtualization) on top of it to give companies the benefits of both worlds by taking a holistic platform approach.

An architecture such as depicted below may be appropriate for many edge options where mixed containerized and virtualized workloads are required and space is at a premium.

Another use case for bare metal that I am seeing is the need to leverage raw compute power that can best be provided by bare metal hardware.

Will this work?

These architectures are certainly not futuristic. They are supported configurations. Red Hat has tested and certified implementing OpenShift on Bare Metal using several different methods. OpenShift Virtualization builds on years of experienced with Red Hat Virtualization. The KubeVirt community is established and maturing.

Conceptually these architectures make a lot of sense, but how do they work in practice?

  • Can bare metal OpenShift be deployed easily and is it stable?
  • Can it scale?
  • What is the cost profile?
  • Can I manage my virtual machines using my existing automation approach?
  • How do you enable high-availability when shared storage is not an option?
  • What do I lose with removing the virtualization layer?

While there are multiple customers running Bare Metal OpenShift in production, I have customers that are exploring running OpenShift on bare metal. I’ve decided to make a commitment to them and flip my home lab from a traditional virtualization platform with OpenShift running in a VM to the Bare Metal Edge architecture described above.

I was curious about what I’d find and the pitfalls I’d come across. To encourage discussion and create a community I decided to start blogging about my successes and failures along my journey.

Listed below are the blogs that I have published on this subject.

--

--