Many enterprises devote as much as half of their data center space and resources to activities such as testing and development. Usage of these test and development environments varies dramatically, though; sometimes they are working flat out, while at other times there are plenty of resources standing idle.
For that reason it often makes sense to move them to the cloud. That way you pay for them only when they are needed, and not when they are idle.
And that's the raison d'etre of Ravello Systems, a company we looked at last year, and one that uses virtualization to encapsulate entire test and development environments — servers, networks and all — and run them in the cloud.
Back then Ravello was in beta testing, but now the company's service is fully up and running. A quick recap: the company was founded by members of the team that introduced the KVM hypervisor (now the standard virtualization technology on Linux), and it used nested virtualization, or "cloud application hypervisors."
Whatever you want to call it, the idea is that you run Ravello's hypervisor, called HVX, on a virtual machine in the cloud, and then on top of that hypervisor you can run VMware or KVM-based virtual machines. In fact, you take an entire application, with all the servers required to run it, encapsulate it and then move it to the cloud to run on the HVX hypervisor, which is itself running on the public cloud provider's hypervisor. Developers can also use Android on HVX to run scale tests of thousands of Android clients.
Ravello's HVX Hypervisor in a Nutshell
It's complicated stuff, but here's an explanation of HVX in a nutshell: It is capable of running unmodified guests on top of already virtualized hardware. Conventional hypervisors such as VMware ESX, KVM and Xen are designed to run on physical x86 hardware and use virtualization extensions offered by modern CPUs (Intel VT and AMD SVM) to virtualize the Intel architecture.
HVX, on the other hand, is a nested hypervisor that runs inside a virtual machine, where these hardware extensions are not normally available. Instead, HVX employs a technology called Binary Translation to implement high-performance virtualization that does not require these virtualization extensions.
Networking is also taken care of when you move an application to the cloud to run on HVX. "You run HVX on cloud virtual machines, and all the HVXs talk to each other and form a networking overlay. That means that you can define any networking topology you like," explains Ravello's Navin Thadani.
Since exiting from beta Ravello has revealed its pricing. It's designed so you pay per application rather than per virtual machine, and it's based on the application's size and complexity, and whether you want to optimize cost or performance. An application that needs 4 vCPUs, 8GB RAM, requires complex networking and that is optimized for lowest cost can be run on Ravello for 0.42 cents per hour, as an example.
Ravello has also extended its cloud support — it has always run on AWS, HP and Rackspace clouds, but now it also runs in the Google Compute Engine cloud.
Customers actually pay their bills to Ravello rather than directly to these cloud providers, which begs the question as to what happens when cloud providers cut their prices — as Google and Amazon both did recently?
"Our prices are connected to the underlying clouds, so if Amazon cuts its prices it lowers our cost structure," says Thadani. Savings are passed on to customers, he explains, "but not 1:1 immediately."
At the moment Ravello is concentrating on helping enterprises move their dev and test environments to the cloud, but Thadani doesn't rule out extending this in the future. "If the cloud prices keep decreasing, then at some point it will be feasible to use Ravello to run an entire production environment in the cloud," he predicts.
Paul Rubens is a technology journalist and contributor to ServerWatch, EnterpriseNetworkingPlanet and EnterpriseMobileToday. He has also covered technology for international newspapers and magazines including The Economist and The Financial Times since 1991.