There is no denying the momentum of the Kubernetes platform and ecosystem, with virtually every enterprise looking to run containers at scale at some stage of adopting it. As a result, the Azure Kubernetes Service (AKS) is now one of the fastest growing services in the history of Azure. However, there is still a lot that can be done to make it easier to use, for newcomers and experts alike.
Today, I am pleased to announce a series of new capabilities and investments that further demonstrate this commitment to making Kubernetes easier.
Azure Kubernetes Service virtual nodes enters public preview
Customers consistently tell us that they love the speed and portability of containers. In fact, they love containers so much that they’d prefer to spend as little time as possible thinking about virtual machines, even when it comes to acting as hosts for their containers. With the launch of Azure Container Instances (ACI) last year, Azure became the first public cloud to offer containers as a top-level resource. Today, we are announcing the public preview of AKS virtual nodes, which enables you to extend the consistent, powerful Kubernetes API provided by AKS with the scalable, container-based compute capacity of ACI. This combination brings many of the benefits of serverless platforms to the rich, consistent API provided by Kubernetes.
With AKS virtual nodes, you can respond to a spike in demand in seconds by allocating precisely the number of additional containers that you need, rather than waiting for additional VM-based nodes to spin up. Since ACI is billed by the second based on the resources that you specify, you can exactly match your costs to your workloads. And because virtual nodes look just like any other node in the cluster, you can continue to use the same tools, processes, and knowledge you do today.
Enabling virtual nodes is as simple as flipping a toggle in the portal:
To get started, check out the documentation for enabling virtual nodes via the portal or the CLI.
Donating the Virtual Kubelet project to the Cloud Native Compute Foundation
AKS virtual nodes is built on the open-source Virtual Kubelet (VK) project. We launched VK at KubeCon 2017 as a way of starting a conversation with the community about how the Kubernetes API could be extended to deploy and manage containers in non-traditional compute environments like ACI. Since then, a number of VK providers have been added, enabling integrations with services like AWS Fargate, Alibaba ECI, and Azure IoT Edge. Today, we are pleased to announce the next step in that journey with the donation of the Virtual Kubelet to the Cloud Native Compute Foundation (CNCF) as a sandbox project. By working within the CNCF, we hope to encourage even more participation in this conversation and further formalize an idea in which there is no shortage of interest.
Azure Container Instances GPU support enters public preview
Making containers truly first-class in Azure means ensuring that they can support all different types of compute workloads, just as virtual machines do. Recently, we added support for ACI containers joining private virtual networks, enabling interaction with other resources in your environment. Today, we are adding GPU support to ACI, enabling a whole new class of compute-intensive applications through AKS virtual nodes. Initially, ACI will support the K80, P100, and V100 GPUs from Nvidia and you will have the ability to specify the type and number of GPUs that you would like for your container. As with CPU and memory, GPUs are charged per second, based on the number of cores requested. Check out the docs to get started!
An ongoing commitment to making Kubernetes easier
Our announcements today represent another demonstration of our commitment to making Kubernetes easier to adopt and use, both for developers and operators. By simplifying how you interact with the Kubernetes platform and expanding the types of compute that you can leverage, we hope you’ll be able to spend more time building your applications and accomplish even more with AKS.
If you’re new to Kubernetes, you might want to check out our new conceptual documentation.