It was magic at first, running multiple VMs on one server using a hypervisor. Then the magic was running VMs on a hypervisor that was itself a VM on another hypervisor. What if I told you that there was a cloud that is built on top of other clouds? A sort of cross-cloud cloud? And this cloud allowed you to take your VMs from vSphere and KVM straight into the cloud? No conversion, just copy the VMs to this new cloud.
Naturally the new cloud is real, it is by a company called Ravello Systems. The team at Ravello also built the KVM hypervisor, they have now built their own hypervisor (HVX) that runs on top of both Amazon’s and Google’s public cloud. The Ravello platform looks like a normal cloud management console, only when you come to deploy new VMs you can choose which public cloud you’d like the VMs to reside on. The HVX hypervisor can provide the same virtual hardware specification as VMware’s ESXi, including VMXNet3 NICs and PVSCSI storage adapters. Your vSphere VMs do not need to be reconfigured and installing VMware tools is recommended to get best performance, just like on vSphere. If your VMs start on KVM then HVX can deliver the VirtIO devices that they are already configured for. This is a great help to build a test environment for testing upgrades of operating systems and software inside VMs. Just clone your production VMs and upload to Ravello.
Ravello has the usual cloud platform service catalogue of VMs and ISOs, but it also has a construct like VMware’s vApps. A Blueprint is a template of a group of VMs, so the group can be deployed at once. You can even deploy multiple isolated copies of a Blueprint to allow concurrent testing. The Ravello platform handles standing up the out cloud platform VMs and the HVX hypervisor, then it sets up a virtual network between groups of nested VMs. The networking allows forwarding of public IP addresses to VMs, either dedicated IPs for VMs or sharing public IPs and translating TCP ports.
Ravello also aggregate cloud charging, you will be charged for the amount of storage and compute your VMs use from the underlying cloud by Ravello. This makes it important that you shut down VMs when you aren’t using them. Happily part of the process of deploying or powering on a VM is defining a drop dead time. This is a time when the VM will be automatically shut down to stop charge getting out of control. Naturally you can set this time to Never, with the resulting potential for a large bill but allowing production VMs to keep running. When you first deploy a group of VMs you get to choose what cloud provider and availability zone you would like to deploy onto. This also affects the performance of your VMs and therefore the cost.
Ravello have a great way for you to test their platform, a free trial. The trial allows you to run a limited set of VMs on their platform for up to a month. This is an easy way to test the waters and see for yourself what the Ravello platform is about. The trial is a month, or a maximum cost for the VMs.
You can read more about what my fellow bloggers have done with and think about the Ravello platform on these page:
- Andrea Mauro – Ravello System and its Lab as a Service solution
- William Lam – Running Nested ESXi / VSAN Home Lab on Ravello
- Mike Preston – A Google Cloud to Amazon vMotion – The Ravello Way!
- Scott Lowe – Running vSphere on AWS or GCEScott Lowe – Running vSphere on AWS or GCE
Last week Ravello announced that their HVX hypervisor now supports running VMware’s ESXi as a nested VM, which can then run it’s own VMs. Now we have three layers of hypervisors, the cloud hypervisor, HVX and ESXi. Definitely a case of inception. This is what I’ve wanted since I first heard about Ravello at the OpenStack summit in Atlanta a year ago. With Nested ESXi support on Ravello it is possible for my AutoLab to run on public cloud. Rather than buying a beefy physical machine to run your vSphere study lab you can rent the capacity from Ravello and just run the lab when you need it. Look for a new build of AutoLab in the next weeks that supports Ravello.