Blog Post

5 takeaways from the CloudStack-OpenStack dustup

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Citrix(s ctsx) stunned its OpenStack partners with news on Tuesday that it is setting up CloudStack as an Apache-based rival to the OpenStack cloud platform.

There’s been a lot of jawing and some bad feelings on both sides: OpenStack wasn’t ready for primetime; Citrix was disloyal, yada yada yada. But now as the smoke starts to clear, here are some key takeaways from the latest cloud spat.

1: The API debate is (still) over

Amazon(s amzn) won. CloudStack’s support of the Amazon API confirms this, much as the Amazon-Eucalyptus deal did a few weeks ago. (Eucalyptus — another OpenStack rival — claims compatibility with the Amazon APIs in a move that should make it easier for private clouds running Eucalyptus to interoperate with workloads in the Amazon public cloud.)  As Forrester Research analyst James Staten blogged:

“Like Eucalyptus, Citrix CloudStack provides API compatibility with Amazon Web Services’ Elastic Compute Cloud (EC2). The OpenStack community is debating the degree of compatibility its software should have with this de facto and clearly proprietary API set. Citrix no longer has to debate this issue or be beholden to the OpenStack group decision here which looks to be leaning away from EC2 compatibility.”

Still, there’s support and then there’s support — and there is real confusion about which clouds support which Amazon APIs.  As Adrian Cockroft, Netflix’ director of architecture of cloud systems, tweeted Tuesday:

Also, there’s considerable disagreement about allowing Amazon to set the API table is really a good thing. But it’s clear it has done so.

2:  Multi-cloud is huge

Despite the furor around whether or not Citrix threw OpenStack under the bus and fractured the open-source cloud alliance against Amazon, the fact is that many companies are weighing or already running multi-cloud solutions. Some do so to mitigate risk, and others because some clouds suit some uses better than others.  And if you’re of that mindset, what’s one more cloud among friends?

Recent research by Rightscale, a company that provides a dashboard and management tools for all the clouds, shows that 87 percent of its customers are deploying multiple clouds, up from 84 percent in September.

3: Rackspace needs to loosen up

Critics say Rackspace has not sufficiently loosened its grip on the OpenStack effort, which it launched with NASA in 2010, and that control is hurting the cause. When Rackspace announced plans for a more open OpenStack Foundation last fall, it pledged to relinquish control — something that critics say it has not yet done.

In the CloudAve blog, cloud expert Krishan Subramanian wrote that CloudStack’s move will force OpenStack to act more forcefully — and quickly.

Now, the OpenStack Foundation will be forced to have a more democratic and more equitable solution to the problem. Any failure to do so will ultimately result in smaller players, who feel less appreciated or ignored, quietly move to the Apache CloudStack ecosystem. I think that this announcement will add considerable pressure on Rackspace and the rest of the OpenStack community to work together to set up the foundation right.

4:  OpenStack needs to get the lead out

CloudStack — which Citrix bought along with for $200 million last year — may well be more mature than OpenStack and have more features, but it has less mindshare. Putting that effort into the Apache realm should make it more palatable for corporate use, since it’s no longer controlled (ahem) by one company. That means OpenStack really needs to get its house in order both in terms of governance and feature parity.

In short, with Eucalyptus back in the game, CloudStack gaining momentum, and OnApp — which enables web hosting companies to federate clouds — signing on more service providers, “the window for OpenStack to become the Linux of IaaS is beginning to close,” wrote Forrester’s Staten. (For more on federated clouds, see this GigaOM Pro report, subscription required.)

5:  OpenStack Spring should be a big show

The OpenStack Conference in a few weeks should be interesting given the recent hubbub. Citrix said one reason it went its own way was that OpenStack development was in the slow lane. Look for big time OpenStack partners — including Hewlett-Packard(s hpq)  — to highlight their OpenStack-based goodies.  Biri Singh, VP and GM of HP’s cloud services unit will be among the keynoters in San Francisco.

Now more than ever, OpenStack needs its big boy customers and partners to step up and show that the infrastructure is ready for prime time, is not under Rackspace’s thumb, and can take on Amazon — and now CloudStack as well.

Photo courtesy of Flickr user

3 Responses to “5 takeaways from the CloudStack-OpenStack dustup”

  1. Jim Plamondon

    The creation of two AWS “zombie projects” doesn’t prove that the AWS API has become the de facto standard in cloud computing. Quite the contrary! It shows the incredible momentum behind OpenStack (and OpenStack’s API), which is causing its private-cloud competitors to run for cover. (Heck, I predict that even VMWare will make a similar announcement before OpenStack’s upcoming conference.)

    What’s a “zombie project”?

    In traditional voodoo, “zombies” are corpses raised by a sorcerer and completely controlled by the sorcerer’s will, with no freedom of thought, choice, or action. If raised soon enough after death, they may LOOK alive, but they are…merely zombies.

    Eucalyptus and CloudStack are now “zombie projects.” They are utterly (although indirectly) controlled by Amazon, through the projects’ need to maintain compatibility with Amazon’s AWS APIs.

    These projects may once have had independent life, but — threatened with economic death by the meteoric rise of OpenStack — they have chosen to become AWS zombies instead. As such, they have lost all freedom of thought, choice, and action. They can implement no new cloud services, other than those previously implemented by AWS. Their architectures cannot be creatively improved, lest they break compatibility with AWS. They are *zombies*, serving only Amazon’s will — and serving only Amazon’s strategic interests. Every company that comes to rely on these zombies becomes infected with their zombie virus, and thereby exposes itself to a future Cloud Zombie Apocalypse.

    In creating these zombies, Amazon is stealing a page from Microsoft’s “How to Build a Monopoly” playbook (which I helped write, back in the 1990’s).

    Consider one of Microsoft’s best-documented zombie-creation programs, the Windows Interface Source Environment (WISE, Microsoft used the WISE program to combat Sun’s WABI & PWI projects, which sought to create Unix-based implementations of Microsoft’s Win32 API. This API was the key to Microsoft’s vendor lock-in, and so Microsoft wanted to control any such implementations.

    By licensing Windows’ source code to third parties such as Hunter, Mainsoft, and Bristol, Microsoft created “zombie projects” that out-competed Sun’s WABI & PWI. Once the tactical threat of WABI & PWI was eliminated, Microsoft slaughtered its zombies, as detailed in the records of subsequent anti-trust actions (see Any apps that targeted the Win32 API could then run only on Windows…which was Microsoft’s strategic goal.

    Microsoft could kill its WISE zombies because of the fine print in the WISE program’s license agreement, which — is it had been public and were closely examined — would have revealed signs of the coming Win32 Zombie Apocalypse.

    At the same time, Microsoft created a zombie project on the Mac: the “Visual Studio 4 Cross-Development Edition for Macintosh,” to further establish its Win32 API as the industry’s de facto standard. This zombie enabled developers to recompile their Win32-based apps to run on the Mac. This worked so well that at least one such app won an “Eddy” award for best new Mac app one year. Indeed, it worked so well that Microsoft’s apps group burdened its license agreement with restrictions on the kinds of apps that could be created with it, to ensure that no one could use it to create an Office-killer.

    Again, the zombie project’s fine print contained signs of the coming Win32 Zombie Apocalypse.

    Zombie-creation is just too effective a tactic NOT to use, if you’re trying to lock the entire industry into a proprietary de facto standard.

    That is, of course, exactly what Amazon is doing today. It is attempting to establish its proprietary AWS API as the industry’s de facto standard, and one of its tactics is the resurrection of corpses like Eucalyptus and CloudStack as zombies. These zombies walk, and talk, and smile, and market just like living projects…but Amazon can — and will! — slaughter them as soon as they have served their tactical purpose.

    Some pundits, seeing Amazon re-animate these private cloud zombies, have taken Amazon’s actions as recanting its leadership’s previous denunciation of private clouds as “false clouds.” Ha! Far from it! The zombies are a mere tactic, advancing Amazon’s strategic objective of providing the One (Public) Cloud to Rule Them All. The *sole* (not “soul,” because zombies have none) purpose of Amazon’s zombies is to slow OpenStack’s momentum. Only by preventing OpenStack from establishing a truly open cloud stack — with an open API and an open implementation, designed through an open process, openly governed — can Amazon establish the kind of Total Industry Domination that Microsoft attained in the 1990’s.

    When Amazon slaughters its zombies, every company that has come to depend on the zombies will realize that they, too, have become infected with the zombie virus. They, too, will have lost their free will. Every line of code that they have written to AWS’ API will compel them to switch from the zombies’ private clouds to Amazon’s public cloud (or to a private cloud service that Amazon will, by then, have started).

    For Amazon, this outcome is game, set, and match. Total Industry Domination. Monopoly. The pleasure of being overseen by the same government agencies that have thus far winked at Wall Street’s ongoing rape of the global economy.

    “Monopoly.” What a lovely word — if you own it yourself!

    If, on the other hand, you wish to deploy your own cloud services, or write applications that use cloud services…then becoming infected with AWS’ zombie virus should be avoided at all costs. Instead, you can inoculate yourself against the coming Cloud Zombie Apocalypse by using OpenStack — an open implementation of an open API, openly designed and openly governed — to meet your cloud computing needs.

    That realization is what has driven the incredible rise of OpenStack (the fastest-growing open source project in history) and — in reaction — the desperate self-zombification of Eucalyptus and CloudStack.

    So: is AWS’ API already the de facto standard cloud computing API? Not even close. It is, at most, the CP/M to OpenStack’s DOS. And as these references show, this competition is still in its *very* early stages.

    Jim Plamondon
    Director, Developer Relations
    Rackspace Cloud

  2. > Now more than ever, OpenStack needs its big boy customers
    > and partners to step up …

    Not, not really, this is just the GigaOm press trying to stir the pot so they have something more to write about!

  3. I find some of the claims in this article to be nothing short of bizarre. First, the claim that Amazon has “won” the API debate completely ignores several facts, including some that are in the article itself. The OpenStack API is not the same as the Amazon API, although it uses some elements, and the article states that they have more mindshare than CloudStack. SoftLayer, GoGrid, OpSource, Linode, CloudSigma, Zunicore (Peer 1) and pretty much everyone else (we track 600 different IaaS providers) has a proprietary API. Further, if you’re talking about multi-cloud or hybrid environments among customers, you can’t ignore vCloud (VMWare), which is the de facto standard within enterprises. Next, you say that RightScale supports “all” the cloud providers, when the article you link to says they only support six.

    I’ll say it again, so that people in the Amazon-centric echo chamber hear it: There are over 600 different IaaS providers worldwide. Many of them have their own API, because they wrap the API of whatever underlying technology they’re using. VMWare and vCloud are not going to go away. The Amazon API is *not* a standard, de facto or otherwise – it’s not even one of two competing standards. API proliferation is increasing.