Azure Container Instances Take VMs Out of the Mix
Microsoft's new Azure Container Instances (ACI) introduces a way to spin up workloads with the precise amount of CPU cores and memory needed on demand, use them for as little as a few seconds and instantly remove them and pay just for what was used. And it does so with the need to provision and manage Hyper-V or any other virtual machines.
ACI, revealed yesterday, is what Microsoft believes is the fastest and easiest way yet to deploy and manage container-based workloads in the cloud, pushing it up the infrastructure layer. Corey Sanders, director of Azure Compute at Microsoft announced the new service during a webcast, where he demonstrated the ability to spin up and take down container-based workloads.
Microsoft's existing Azure Container Service, released over a year ago, introduced the ability to deploy containers in its public cloud service. Using a choice of orchestration tools such as Mesos DC/OS or Docker Swarm, the current Azure Container Service deploys containers as a set of virtual machines. Administrators must manage both the VMs and containers, whereas the new ACI is able to provide that on-demand provisioning and deprovisioning capability without having to use virtual machines to spin up the containers.
"That is what makes Azure Container Instances unique," Sanders said. "It directly deploys the containers instead of deploying and requiring customers to manage the virtual machines underneath. This pushes the infrastructure up a layer, enabling customers to just work with containers, and no longer have to worry about container deletion, creation, patching and scaling of virtual machines. That's all taken care of by just exposing the container as the top-level resources."
Does this new capability portend the demise of the virtual machine? Given Hyper-V is a key component for which Azure itself is built upon, probably not anytime soon. Even with ACI, Sanders noted that Docker Hub and the Azure Container Registry will be supported and that Hyper-V virtualization in Azure enables the deployment and secure isolation of containers. It's still early days for containers.
"Containers are certainly emerging as a platform for newer workloads and for enabling some of the points around agility and scale," he said. "[But] we expect that many workloads will require the isolation characteristics and scale characteristics that come with full virtual machines. We do expect those will continue to be used, and deployed by customers, both in the public cloud and on premises, whether it be on Hyper-V or other hypervisors."
Posted by Jeffrey Schwartz on 07/27/2017 at 2:20 PM