On Thursday, Google announced that it was bringing nested virtualization to the Google Compute platform in beta, offering new options for virtual machines in the cloud.
To best understand what this practically means for customers looking to leverage the service, it’s best to break the topic down to the three basic questions of what, why, and how.
1. What is nested virtualization?
Nested virtualization is the technique of running a hypervisor inside another hypervisor. Think of running Hyper-V within VMware ESXi. In this example, the Hyper-V instance could host regular workloads such as Windows Server 2016 running a web server.
SEE: Quick glossary: Virtualization (Tech Pro Research)
The ability is brought about by way of improvements in x86 processors. With the goal of increasing virtualization support and performance, both AMD and Intel created virtualization-specific extensions. As an after effect, nested virtualization is possible with the performance overhead mitigated by improvements in the extensions.
2. Why use nested virtualization?
There are many use cases for nested virtualization. I’ve been using nested virtualization in my home lab for years. A beefy workstation with 32GB of RAM running VMware Workstation allows me to run three ESXi hosts, each with 8GB of RAM. Three ESXi hosts are the bare minimum for running a vSphere lab.
A growing use case for more critical workloads involves the public cloud. Oracle Ravello, for example, has long provided the capability to run nested virtual machines in both Google Cloud and Amazon Web Services (AWS). Google direct support is a validation of the need to run unmodified workloads directly on the cloud. An alternative to VMware Cloud on AWS is to run ESXi directly on Google Compute Engine. While not supported by VMware, the solution provides a hack for getting test and dev workloads on elastic compute resources.
3. How to use the service
You can enable the service on new or existing instances. Scale Computing recently showed one of the more interesting use cases. At Tech Field Day in San Francisco, Scale Computing announced the general availability of their HC3 platform on Google Compute. HC3 is Scale Computing’s virtualization platform built on KVM. Existing customers running an HC3 system now can replicate VMs to a nested HC3 system in Google Compute Engine.
The HC3 service is initially meant to provide disaster recover capability. According to Scale Computing CEO Jeff Ready, the largest Scale customer is migrating from 30,000 HPE servers, but the majority of Scale Computing customers are SMBs. The solution provides a simple way to provide disaster recovery without having a large IT staff to design and maintain the solution.