Marten Mickos – the former MySQL chief executive who now heads build-you-own-cloud outfit Eucalyptus Systems – has defended the Eucalyptus platform against recent criticism of both its "open core" model and its ability to scale beyond a relatively small number of servers.
His comments come as Eucalyptus releases a new incarnation of its open source platform, which exists alongside the company's for-pay enterprise platform. Version 2.0 of the open source Eucalyptus is designed to, yes, improve the platform's ability to scale.
"We have improved scalability on the front-end and the back-end," Mickos says. "We aim to improve transactional scalability as well as resource scalability." The new version also supports virtio, a virtual I/O Linux framework, and iSCSI targets for Elastic Block Storeage (EBS) volumes, a much requested feature.
This summer, NASA announced that its Nebula infrastructure cloud – a means of offering scalable, on-demand processing power across NASA and eventually other federal agencies – will abandon its Eucalyptus base in favor of a home-grown platform known as Nova. According to chief technology officer Chris Kemp, NASA is moving away from Eucalyptus because the platform didn't scale as well as the agency would like – and because his engineers weren't able to submit patches to the open source Eucalyptus project that would have improved its ability to scale.
Eucalyptus – an attempt to mimic Amazon's EC2 infrastructure cloud inside private data centers – was created as an open source project at the University of California, Santa Barbara. But after its founders took the project commercial with the formation of Eucalyptus Systems, they adopted what Mickos calls an "open core" business model. There's still an open source version of the platform. But there's also an enterprise version that incorporates proprietary software.
NASA wants to build an infrastructure cloud that spans one million physical machines and 60 million virtual servers, and according to Kemp the open source Eucalyptus just wasn't suited to such a project. "With the architecture of the open source code in Eucalyptus, by our team's analysis, you couldn't get close to that," he told us.
"Part of that is that it's a product. It's not a framework. If you need a faster queuing service or a faster database engine, all of that stuff is integrated directly into the Eucalyptus product. You can't easily pull it apart and replace the queuing engine or networking engine with other systems that scale better."
Rather than fork Eucalyptus, NASA built its own compute engine and fabric controller, Nova, and this has now been open sourced as part of the OpenStack project.
The scuttlebutt around the industry has long been that Eucalyptus doesn't scale as well as it should. But Mickos says that such rumors should be laid to rest – though he acknowledges that the platform may not be appropriate for the sort of epic cloud NASA intends to build. Eucalyptus is meant for use within the enterprise.
"I welcome the challenge [from critics]," Mickos tells us. "In early versions, we scaled less than we do now." He points out that in the early days – around version 1.1.2 of the platform – Eucalyptus offered a quick download that ran many critical services on the same physical server.
"This made it very easy to download and very easy to get going," he says. "But when a cloud scales, you need to move the cloud controllers and the cluster controller and other services on different machines, and we didn't make this happen automatically. Maybe this was a disfavor to our customers, but they mistakenly put things on the same machine that shouldn't have been."
Mickos says that a million-machine cloud along the lines of what NASA wants to build is not "specifically in [Eucalyptus's] focus area." But he points out that NASA has yet to remove Eucalyptus from its production machines, and he claims that in many areas, Eucalyptus exceeds the performance of Amazon's EC2, the public infrastructure cloud that provided the inspiration for the platform.
He also disagrees with NASA's claims about the agency's inability to get new patches into the open source version of Eucalyptus. "I think there has been a misunderstanding somewhere along the line," he says. "We do accept contributions. We love them. We don't mind if the contributions relate to what we have in the enterprise version."
He says that NASA's problems related to copyright issues. Eucalyptus asks for copyright transfer when users contribute code, and Mickos claims that NASA was unwilling to do so because of certain government regulations. "I believe that was the sticking point," he says.
But Mickos adds that he is merely speaking on behalf of others. And his account is quite different from Kemp's – and from others with knowledge of the situation who have spoken to The Reg.
Asked a second time about Kemp's claims that NASA was unable to contribute patches because they conflicted with the enterprise version of Eucalyptus, Mickos offers a mea culpa, but NASA's story doesn't reflect the Eucalyptus open source project as a whole.
"I apologize if NASA believes that was the case and for any faulty communication on our part," he says. "We have a specific, well-crafted contribution mechanism and agreement that we are following, and for many, if not most, it is OK."
Version 2.0 of the open source Eucalyptus is now available. The platform is bundled with Canonical's Ubuntu Linux. Thus, the "Karmic Koala" nickname attached to Ubuntu 9.10. ®