Ravello shifts virtualization from servers to clouds
Laurent Lachal, Senior Analyst, Software – IT Solutions, OvumThe infrastructure-as-a-service (IaaS) market is growing up. Vendors are endeavoring not only to help enterprises define, configure, and deploy increasingly complex applications, but also to abstract these applications from the underlying IaaS infrastructure. It is shifting from a bottom-up infrastructure-centric view to a top-down application-centric view.
This is happening in a variety of ways, including Ravello Systems’ Cloud Application Hypervisor (CAH), whose open beta launched in February 2013, the same month Ravello raised an additional $15m. Under its “Develop in the cloud, Deploy anywhere” tagline, Ravello is initially targeting the public cloud-based development and testing (dev/test) market.
More specifically, it is focusing on enterprises that wish to move existing as well as new applications to public clouds for dev/test purposes without having to make any changes to the applications. In its report entitled “On the radar: Ravello Systems”, Ovum explains why enterprises should take a look at the new technology that is shifting the focus of virtualization from servers to the cloud.
From server to cloud virtualization
Standard hypervisors abstract single virtual machines (VMs) from underlying physical servers and package them as monolithic entities that combine executables, data, and configuration files. Ravello’s software-as-a-service (SaaS) Cloud Application Hypervisor (CAH) abstracts multi-VM applications packages called capsules from the underlying data center infrastructure.
CAH relies on HVX, which is a nested hypervisor with software-defined network and storage capabilities, that runs capsules defined via a graphical application framework (AP) design tool. Users upload their VMs and use the AP’s drag-and-drop graphical interface (or CAH’s API) to define an application’s architecture and environment as a capsule blueprint. They can also use Ravello-provided blueprints. They can then instantiate one or more version of the capsule to either a standard data center, a private cloud, or a public cloud.
A mix of strengths and weaknesses
CAH removes the need to change (translate, rebuild, or reconfigure) existing applications. Users can define the networking configuration to be exactly the same as that in the data center. CAH takes care of VM/application stack provisioning, configuration, and cloning, with no need to learn the variety of approaches taken by the different underlying public clouds.
Developers and test engineers (employees and third-party contractors) can get their own application instance, which helps improve code quality, encourages experimentation, and reduces time to market. On the other hand, CAH is only available in beta form with (negotiable) restrictions.
Its functionality needs to expand. For example, it should make it as easy for IT people to quickly spin off their own dev/test environment as it is for IT managers to keep track of who does what. This is not yet the case. The performance overhead of its nested hypervisor ranges from less than 10% on CPU-bound applications, up to 20% to 40% on I/O-bound applications compared to 4% or less for standard KVM or Xen hypervisors. This is one of the reasons why Ravello is targeting dev/test, rather than production, usage scenarios.
More to come, and until general availability it is free to test
The ambition of the open beta is to bring CAH to the attention of the enterprise market at large. It is totally free (Ravello pays for everything including VM costs). General availability (GA) is planned for mid-2013, at which point usage-based pricing, including a free startup tier, will kick-in. Next, Ravello plans to develop CAH in a variety of directions including more public cloud support (beyond the Amazon Web Services, HP, and Rackspace clouds it currently supports), additional functionality such as team collaboration functionality, and more out-of-the-box customizable VMs/Blueprints, including ones that support auto-scaling, regional high availability, and eventually even multi-cloud high availability.