What Ubuntu’s Move to OpenStack Means for Eucalyptus

This week, the Ubuntu project announced that “future versions of Ubuntu Cloud will use OpenStack as a foundation technology,” apparently replacing Eucalyptus as the enabling infrastructure behind Ubuntu’s cloud capabilities. Eucalyptus does not (yet) appear to have been dropped completely, and existing deployments will definitely be supported for the next few years. But this is the latest in a series of setbacks for the platform, and it is becoming increasingly unclear whether it has a compelling future.

Version 1.0 of Eucalyptus‘ open-source cloud infrastructure solution emerged from the University of California, Santa Barbara in May 2008; it went on to power NASA’s innovative Nebula Cloud, ship globally alongside Ubuntu Linux and on Dell servers, and launch over 25,000 clouds of all sizes. Then NASA looked elsewhere for its cloud computing solution, Ubuntu got increasingly close to OpenStack and the whole picture began to look a lot less favorable as Eucalyptus lost out again and again to OpenStack and others.

It is, however, important to distinguish the Eucalyptus project from the commercial Eucalyptus Systems. The latter sells services around the core software. That particular business — headed by former MySQL boss Marten Mickos and Eucalyptus project lead Rich Wolski — should continue to offer a viable set of services to companies requiring a cloud-computing solution, regardless of the other channels by which the project code itself may or may not be made available for download.

Eucalyptus benefited from the outset because it was designed around an API that performed almost identically to Amazon’s. Organizations were able to test applications on Eucalyptus and then push them live on Amazon Web Services, or vice versa. They could also run core applications in-house on Eucalyptus, and draw upon additional compute capacity from Amazon when required without having to invest in significant rewriting of their code. As Amazon Web Services grew in reach and ubiquity, Eucalyptus should have been well positioned to address the hybrid- and private-cloud markets.

Partnerships such as those with Ubuntu and Dell should have made Eucalyptus difficult to miss for anyone seeking their own cloud infrastructure solution, but Eucalyptus consistently failed to attract lasting and widespread attention of the sort that OpenStack did almost from the moment it launched in July of last year. OpenStack has signed up over 60 supporting companies including Microsoft, GigaSpaces and Cisco. It is replacing the code beneath the Rackspace Cloud product and powering private-cloud installations at big enterprise customers such as AT&T.

This week’s announcement sees OpenStack replace Eucalyptus as an integral of Ubuntu Cloud. Although it is made clear that Eucalyptus “will be supported by Canonical,” the announcement appears deliberately vague about how easy it will be for future customers to choose between OpenStack and Eucalyptus. OpenStack is described as a “core part,” while Eucalyptus merely “will continue to be available for download.”

Joe Panettieri of Talkin’Cloud suggests that Canonical is “trying to be tactful” in not issuing an announcement that explicitly dumps Eucalyptus in favor of OpenStack. The ambiguity, however, is not helpful. Over at CloudAve, Krishnan Subramanian describes the announcement as the finalization of Canonical’s “slow divorce” from Eucalyptus, and welcomes it as a “smart move.” At Eucalyptus Systems, CEO Mårten Mickos puts a brave face on the announcement, expressing disappointment with the decision while promising continued support for those areas of the Ubuntu codebase that Canonical is phasing out.

Smart move or not, rejection by the Ubuntu project is a clear blow to Eucalyptus’ credibility. OpenStack’s rise has been given yet another significant boost, and the Eucalyptus project must surely be left wondering where its future might lie. Is it worth continuing to develop code for a shrinking niche market, or is there more value to be gained in porting the best of Eucalyptus’ ideas to OpenStack? Eucalyptus is used today by big customers like Puma, and much rests on whether or not Mickos and his team can persuade those customers to stick with the company. Whether as a provider of professional services around Eucalyptus or as a competitor to OpenStack’s Cloud Builders, there will still be plenty of work for Eucalyptus Systems to do, even if it makes the brave — but painful — decision to let its own baby fade into cloud-computing history.

Question of the week

Should the open source community concentrate on just one cloud infrastructure solution, or continue to develop several?
Relevant Analyst
PaulMiller-med5f09011b40a21c20b941ac1948ceee31-avatar2

Paul Miller

Founder The Cloud of Data

Do you want to speak with Paul Miller about this topic?

Learn More
You must be logged in to post a comment.
5 Comments Subscribers to comment
  1. Marten Mickos Thursday, May 12, 2011

    Thanks for writing about us, Paul. Wow – that’s a grim outlook you are giving us!

    I am seeing an entirely different world from where I am. We have seen 25,000 Eucalyptus cloud start up all over the world, and growth continues. We have production installations dating back 2 years by now. We are in generation 2 of our GA’d product, and this summer you will see the third major release. Puma, InterContinental Hotels, Aeorspace Corporation, USDA, NSA and others are happy users with expanding clouds. Our business is growing faster than before and faster than planned.

    We face competition in the market, but this is mainly from VMware’s vCloud Director and from Abiquo and Cloud.com (and sometimes from CA). OpenStack is far from production readiness. Hardening software for mission-critical production use is hard.

    As all of us private cloud software vendors race into the future, it will be interesting to see where each one ends up. We think we are stronger than ever before, and moving faster. But apparently we have been unable to convey that picture to everyone, for which I apologize. I will take it as my duty to do a better job presenting and representing Eucalyptus. In the meantime, please check out our website for webinars, videos and other interesting content, which we are now adding faster than before.

    Thanks for listening and with kind regards,

    Marten Mickos
    CEO, Eucalyptus
    marten@eucalyptus.com
    http://twitter.com/martenmickos

    1. Marten,

      Many thanks for a thoughtful and informative response.

      I should stress the points I made about the apparent viability of Eucalyptus *Systems* – there seems to be clear scope to offer a range of professional services around the deployment and configuration of private clouds, whether based upon Eucalyptus or some other code line.

      Eucalyptus itself was – and remains – impressive. I remember hearing about it originally from Rich, and have continued to see it as an interesting and viable technological solution to the delivery of private clouds.

      The real issue, then, may be the one to which you refer in your response; despite a viable model and fit-for-purpose technology, others (and OpenStack must feature highly here) have more successfully captured the imagination. Whether that is simply within the press/analyst/clouderati community or more broadly amongst real customers is something we might usefully explore in more depth.

      I shall continue to watch Eucalyptus with interest, and look forward to you being able to put my concerns at rest with a series of successes in the coming months.

      Paul

  2. As a Eucalyptus customer, I’ll confess to some bias on this; we’ve invested a fair bit in rolling the enterprise offering out to support our automation infrastructure for build and QA. I care a lot that it keeps working.

    With that disclaimer in place:

    The fact that the Euca APIs are the same as AWS was a crucial selling point for us. We started on AWS entirely and then built out server capacity inside the firewall. Steady state, we’ve got the capacity we need to handle automatic build/test on Eucalyptus. The ability to lay our hands on physical servers matters to us, because Hadoop needs that sometimes. Because we use the AWS APIs, we can burst out to the public cloud and run on EC2 when the workload spikes. It’s an ideal combination.

    Equally, access to the enterprise software and support package has been important.

    I think it’s great that alternative cloud infrastructure layers are evolving right now. It’s too early for the industry to settle on just one. Canonical is pushing hard to grow its datacenter and server business in a market that has really been defined by Red Hat. Canonical’s cloud offering is obviously a key part of that strategy. I like the fact that there’s a vibrant community around OpenStack, but the Ubuntu endorsement notwithstanding, it’s way too early to nominate a single winner among the cloud abstraction layers. We need more years of experience before that happens.

    Right now, I still like Euca best. It’s the most mature project, for sure, and the commercial focus of the vendor behind it is driving the platform forward at a pace and in a direction that lets us plan the future of the systems we run it on. AWS is a good choice for APIs.

    Mike Olson
    CEO, Cloudera
    @mikeolson

    1. Mike

      Thanks for that – it’s always useful to get insight from those who’ve been making this stuff work for them for real.

      I agree that the API compatibility with Amazon is a key feature, and said so in the post. I am actually surprised that more people don’t use it in exactly the way you describe. It seems an obvious and valuable mode of working to me, and one that surely should ease some of the public/hybrid/private transitions that potential beneficiaries of the cloud occasionally get so agitated about.

      Paul

Latest Research

Latest Webinars

Want to conduct your own Webinar?
Learn More

Learn about our services or Contact us: Email / 800-906-8098