Run your Arm workloads on Google Kubernetes Engine with Tau T2A VMs

At Google Kubernetes Engine (GKE), we obsess over customer success. One major way we continue to meet the evolving demands of our customers is by driving innovations on the underlying compute infrastructure. We are excited to now give our customers the ability to run their containerized workloads using the Arm® architecture. 

Earlier today, we announced Google Cloud’s virtual machines (VMs) based on the Arm architecture on Compute Engine. Called Tau T2A, these VMs are the newest addition to the Tau VM family that offers VMs optimized for cost-effective performance for scale-out workloads. 

We are also thrilled to announce that you can run your containerized workloads on the Arm architecture using GKE. Arm nodes come packed with the key GKE features you love on the x86 architecture, including the ability to run in GKE Autopilot mode for a hands-off experience, or on GKE Standard clusters where you manage your own node pools. See the ‘Key GKE features’ below for more details.

"The new Arm-based T2A virtual machines (VMs) supported on the Google Kubernetes Engine (GKE) are providing cloud customers with the higher performance and energy efficient options required to run their modern containerized workloads. The Arm engineering team has collaborated on Kubernetes CI/CD enablement and we look forward to seeing the ease-of-use and ecosystem support that comes with Arm support on GKE."– Bhumik Patel, Director of Software Ecosystem Development, Infrastructure Line of Business, Arm.

Starting today, Google Cloud customers and developers can run their Arm workloads on GKE in General Availability¹ (updated October 5, 2022) by selecting a T2A machine shape during cluster or node pool creation either through gcloud or the Google Cloud console. Check out our tutorial video to get started

Some of our customers who had early access to T2A VMs highlighted the ease of use in working with their Arm workloads on GKE.

"Arcules offers cloud-based video surveillance as a service for multi-site customers that’s easy-to-use, scalable, and reliable - all within an open platform and supported by customer service that truly cares. We are excited to run our workloads using Arm-based T2A VMs with Google Kubernetes Engine (GKE). We were thoroughly impressed by how easily we could provision Arm nodes on a GKE cluster independently and alongside x86-based nodes. We believe that this multi-processor architecture will help us reduce costs while providing a better experience for our customers."—Benjamin Rowe, Cloud and Security Architect, Arcules

Key GKE features supported with Arm-based VMs

While the T2A is Google Cloud’s first VM based on the Arm architecture, we’ve ensured that it comes with support for some of the most critical GKE features — with more on the way. 

• Arm Pods on GKE Autopilot - Arm workloads can be easily deployed on Autopilot with GKE version 1.24.1-gke.1400 or later in supported regions¹  by specifying both the scale-out compute class , and the Arm architecture using node selectors or node affinity. See the docs for an example Arm workload deployment on Autopilot.
  

• Ease-of-use in creating GKE nodes - You can provision Arm nodes with GKE version 1.24 or later using the Container-optimized OS (COS) with containerd node image and selecting the T2A machine series. In other words, GKE automatically provisions the correct node image to be compatible with your choice of x86 or Arm machine series. 

• Multi-architecture clusters - GKE clusters support scheduling workloads on multiple compute (x86 and Arm) architectures. A single cluster can either have only x86 nodes, only Arm nodes, or a combination of both x86 and Arm nodes. You can even run the same workloads on both architectures in order to evaluate the optimal architecture for your workloads.

• Networking and security features - Arm nodes support the latest in GKE networking features such as GKE Dataplane V2 and creating and enforcing a GKE network policy. GKE’s security features such as workload identity and shielded nodes are also supported on Arm nodes.

• Scalability features - When running your Arm workloads, you can use GKE’s best-in-class scalability features such as cluster autoscaler (CA), node auto provisioning (NAP), and horizontal and vertical pod autoscaling (HPA / VPA).

• Support for Spot VMs - GKE supports T2A Spot VMs out-of-the-box to help save costs on fault-tolerant workloads. 

Enhanced developer tools

We’ve updated many popular Google Cloud developer tools to let you create containerized workloads that run on GKE nodes with both Arm and x86 architectures, simplifying the transition to developing for Arm or multi-architecture GKE clusters. 

When using Cloud Code IDE extensions or Skaffold on the command line, you can build Arm containers locally using Dockerfiles, Jib, or Ko, then iteratively run and debug your applications on GKE. With Cloud Code and Skaffold, building locally for GKE works automatically regardless of whether you’re developing on an x86- or Arm-based machine. 

Whether you build Arm or multi-architecture images, Artifact Registry can be used to securely store and manage your build artifacts before deploying them. 

If you develop on Arm-based local workstations, you can use Minikube to emulate GKE clusters with Arm nodes locally while taking advantage of simplified authentication with Google Cloud using the gcp-auth addon. 

Finally, Google Cloud Deploy makes it easy to set up continuous delivery to Arm and multi-architecture GKE clusters just like it does with x86 GKE clusters. Updating a pipeline for these Arm-inclusive clusters is as simple as pointing your Google Cloud Deploy pipeline to an image registry with the appropriate architecture image. 

A robust DevOps, security, and observability ecosystem

We’ve also partnered with leading CI/CD, observability, and security ISVs to ensure that our partner solutions and tooling are compatible with Arm workloads on GKE. You can use the following partner solutions to run your Arm workloads on GKE straight out-of-the-box.

Datadog provides comprehensive visibility into all your containerized apps running on GKE by collecting metrics, logs and traces to help to surface performance issues and provide context when troubleshooting. Starting today, you can use Datadog when running your Arm workloads on GKE. Learn more.

Dynatrace uses its software intelligence platform to track the availability, health and utilization of applications running on GKE, thereby helping surface anomalies and determine their root causes. You can now use these features of Dynatrace with GKE Arm nodes. Learn more.

Palo Alto Networks’ Prisma Cloud Daemonset Defenders enforce security policies for your cloud workloads, while Prisma Cloud Radar displays a comprehensive visualization of your GKE clusters as well as the containers and nodes, so you can easily identify risks and investigate incidents. Use Prisma Cloud Daemonset Defenders with GKE Arm nodes for enhanced cloud workload security. Learn more.

Splunk Observability Cloud provides developers and operators with deep visibility into the composition, state, and ongoing issues within a cluster. You can now use Splunk Observability Cloud when running your Arm workloads on GKE. Learn more.

Agones is an open source platform built on top of Kubernetes that helps you deploy, host, scale, and orchestrate dedicated game servers for large scale multiplayer games. Through a combination of efforts from the community and Google Cloud, Agones now supports the Arm architecture starting with the 1.24.0 release of Agones. Learn more. 

Try out GKE Arm today

To help you make the most of your experience with GKE Arm nodes, we are providing guides to help you with learning more about Arm workloads on GKE, creating clusters and node pools with Arm nodes, building multi-arch images for Arm workloads, and preparing an Arm workload for deployment to your GKE cluster. 

To get started with running Arm workloads on GKE, check out the tutorial video.