What can I say? We started building servers on top of servers and it temporarily blew people’s minds. The next logical step is to build a cloud inside a cloud. Ravello Systems is trying to make this process simple and easy. Ravello Systems was kind enough to present at Virtualization Field Day 5 in Boston at the end of June and I’m happy that I was able to participate at a delegate. They presented some really fun technology.
All travel expenses and incidentals were paid for by Gestalt IT to attend Virtualization Field Day 5. This was the only compensation given and did not influence the content of this article.
So What is it?
The short answer is Ravello Systems allows you to deploy your private cloud and deploy it on either Google Cloud or Amazon EC2. Big Whoop right? We can spin up virtual machines on Google Cloud or AWS by ourselves, why do we need Ravello Systems in the middle? Well, Ravello makes all of this provisioning simple and easy to do and can create cloud templates right from your vCenter. Ravello allows you to import a VM or template right from your vCenter server and add it to a catalog. From there, you can add more VMs or appliances (like load balancers) to build a cloud blueprint via Ravello’s “Visio” style drag and drop interface. These blueprints can be published to either Google Cloud or AWS to be run and the great thing is that you can redeploy your multi-tier apps over and over again from the Ravello user interface.
How does it work?
In order to provide these repeatable cloud blueprints, Ravello lays down their own hypervisor (HVX) on top of EC2 or Google Cloud. When you deploy a cloud blueprint, Ravello’s decision engine kicks in to decide how many resources need to be provisioned and will deploy their HVX on top of the existing public cloud’s hypervisor, and then lays down your blueprint on top of that. This obviously creates some technical challenges such as how network routing will work etc.
One question you might have is how much of an effect on the price or performance does that extra hypervisor cost you? Ravello claims that a single vCPU on their hypervisor is 98%-99% as good as running directly on the cloud provider’s hypervsior. Memory drops to 80% as good, and the network is 80% as well with one exception. If the VM to VM traffic is all on a single HVX instance, the network utilization is up to 2x faster because it never has to leave the host.
There are a few use cases for their service and just like virtualization was strictly for Test/Dev in the beginning, this is a good use case for Ravello to start as well. You can recreate your entire vSphere environment on top of a public cloud if you want to. Sales demos are another one since an entire solution can be rebuilt from anywhere an internet connection exists.
To me, the best use case is for training. How many times have you been to a training class only to find out that the lab they have you working in is really underpowered due to resource constraints. Now the entire lab can be built on a public cloud provider and destroyed at the end of the week.
Home Labs would fit into this training category as well. ESXi is another image that can be deployed on the Ravello solution. This allows you to test new VMware features without having any new hardware. Fellow VFD5 delegate Mike Preston used Ravello to test out long distance vMotion from his home lab to a Public Cloud nested version of vSphere 6. If you’re trying to understand how this works, just think ESXi on top of HVX on top of (public cloud virtualization layer).
Ravello has a pretty neat solution and some cool technology behind it. I love the idea of using this to build a lab in the Public cloud to cut down on capital expenditures and if you’re a vExpert, you can get started too for free. Ravello is offering 1000 hours for free to test out the solution. Go get your evaluation today.