Hadoop and OpenStack shack up with Project Savanna
Red Hat, Hortonworks and Mirantis are teaming up to turn Hadoop into a first-class customer on cloud platform.
Two of the biggest open-source projects around are about to get intimate thanks to Project Savanna, a new OpenStack component being co-developed by Red Hat, Hortonworks and Mirantis. Savanna is designed to make Hadoop a first-class customer on OpenStack, with the cross-industry team-up announced yesterday at OpenStack Summit.
Just as regular VMs can be spun up via OpenStack’s web-based dashboard, Savanna allows Hadoop instances to be provisioned quickly. Parameters such as cluster topology are configurable via drop-down menus, while worker nodes can be added or removed to scale up and down. In addition, Savanna can direct any unused compute power within an OpenStack infrastructure pool to Hadoop.
Amazon Elastic MapReduce is apparently a direct point of comparison, with Savanna’s distinguishing features being open source APIs and the ability to easily port apps between public and private clouds.
Originally developed by OpenStack consultants Mirantis, Red Hat and Hortonworks will be contributing engineers to the project, and each have their own reasons for getting involved.
As the most prolific contributor to the ‘Grizzly’ codebase, Red Hat is betting heavily on OpenStack and is already trialling its own distribution. The company last month revealed that their big data strategy – to not produce their own Hadoop distribution, but build a “robust network of ecosystem and enterprise integration partners”.
Essentially this boils down to making their own software a desirable platform for Hadoop – and involvement in Project Savanna is the first sign that these claims were more than hot air.
Hortonworks and Mirantis, meanwhile, could benefit from the arrangement if devs can be lured away from locked-in big-data-as-a-service offerings as Amazon EMR or Continuuity.
However, Project Savanna is by no means a cabal between the three – like the rest of OpenStack, it’s under an Apache 2.0 license, and it would come as no surprise if more companies jumped on board. There are also plans for the component to support multiple Hadoop distributions.
For those interested in trying out the 0.1 release of Savannah, Mirantis have put together a quickstart guide, and the three companies will be showing off their wares further at Hadoop Summit in June.
Photo by Nicholas A. Tonelli.