The nice thing about a conference like GigaOM Structure is that everybody who’s anybody in the cloud shows up. Representatives from just about every major vendor were there to talk about their vision for the future of cloud services.

But two sessions above all generated buzz and lunchtime discussion, and not coincidentally both were on the topic of open source clouds, as thought leaders in the space (representing OpenStack, Eucalyptus Systems and Apache CloudStack) passionately took to defending their own approach to ending lock-in and boosting customer choice.

Amazon EC2 API: The Clone Wars

The first, a session on the folly of attempting to clone the Amazon EC2 API by Rackspace Hosting CEO Lew Moorman, caused many, many Twitter arguments to break out. Moorman’s argument essentially boils down to the concept that Amazon owns the EC2 API, which is the de facto cloud standard, and that anyone who tries to imitate it just isn’t going to get to a place of robust functionality.

And for developers, there’s a data gravity problem: If you’re really taking advantage of everything the cloud has to offer (and not merely running legacy applications on top of server instances), your applications are hooking into really deep features of the platform.

So even though technically, your applications are portable, it’s just another type of lock-in. And Amazon’s keeping the best stuff for itself. For its business model, he says, that’s exactly the right move. But it means that providers like Eucalyptus Systems, to use his example, can’t achieve real feature parity.

But when AWS customer Netflix –  an example oddly brought up over and over during Structure – wrote a blog post asking the world to build an Amazon EC2 competitor that scales, Moorman says that it shows the need for some kind of alternative: OpenStack.

Co-founded by Rackspace and NASA, the OpenStack standard provides a fully open source, customizable cloud that’s not limited by adherence to somebody else’s API. And as of August 1st, Rackspace will be moving its entire services portfolio onto an OpenStack-based cloud.

And, as Moorman mentioned later, there’s not much reason you wouldn’t want a cloud to compete with Amazon. Competition puts pressure on the dominant power to continually up its game, and even if you only ever use Amazon Web Services or a cloud with an EC2-compatible API, that can only be a good thing.

OpenStack, CloudStack and Eucalyptus: APIs are only the tip of the iceberg

The other panel that drew plenty of reaction was the discussion over a cloud API standard, where Eucalyptus Systems CEO Marten Mickos, Citrix VP of Cloud Platforms Sameer Dholakia (representing Apache CloudStack) and Chris Kemp, CEO of OpenStack-based private cloud developer Nebula heatedly debated the relative merits of sticking with the Amazon EC2 API.

As our own Alex Williams has noted, the real division during the debate was pragmatism versus a vision for a federated cloud future.

In other words, something that Citrix’s Dholakia and Eucalyptus’ Mickos were totally on the same page with is the idea that compatibility with Amazon EC2 is absolutely critical given that 30 percent of cloud developers have adopted the standard. Meanwhile, Kemp argued that “a de facto standard is not a standard,” and that it’s even more important to look to the future and work hard to create an alternative that levels the playing field and enables real innovation in the service provider space.

InformationWeek has a deep blow-by-blow of the debate, complete with sarcasm and snark, for the curious – including Mickos’ now-infamous rebuttal to Kemp that if Amazon Web Services is the Wal-Mart of the cloud, OpenStack is the Soviet Union.

Following Up: Eucalyptus and CloudStack

After the panel, I was fortunate enough to be able to meet with Dholakia, Kemp and Mickos and discuss their feelings on the matter of cloud standardization and market positioning in more depth than the panel format allowed.

Even one-on-one, Eucalyptus and Citrix sounded a lot alike. Citrix’s Dholakia and Eucalyptus’ Mickos individually reaffirmed that their two cloud platforms are aligned in vision, if not technology.

On the subject of OpenStack’s much-vaunted developer community, Dholakia dismisses the notion that more is better – if OpenStack weren’t so immature and didn’t have so many things to work on, it wouldn’t need so many.

“Yeah, because it works,” Dholakia said in response to the common observation that the most recent release of CloudStack had fewer developers than the most recent release of OpenStack.

Similarly, Mickos says that code contributions aren’t a real sign of the viability of a cloud platform, and Eucalyptus has plenty of developers in the ecosystem creating products and extensions on top of or around the platform, which officially goes fully open source next week.

When asked about why OpenStack has succeeded in capturing the imagination of the cloud computing space where CloudStack has failed, Dholakia explained his belief that it’s simply the platform’s roots as a commercial product that have kept it from the open source spotlight.

But that’s also a strength. CloudStack is provably enterprise ready because it’s been deployed in the enterprise, while OpenStack is only now starting to see anybody take on the still-considerable challenge of deployment. And thanks to the still relatively new stewardship of the Apache Software Foundation, CloudStack is attracting a whole new base of developers. As Mickos pointed out, too, many members of the OpenStack community are also working with CloudStack or Eucalyptus.

Another way Eucalyptus and CloudStack are in alignment is that each in its own way attempts to hedge against a VMware monopoly in the cloud computing space. VMware has something like 60% the hypervisor, which is the “gateway” to the cloud, as Mickos puts it. A completely VMware-dominated ecosystem would sentence enterprises everywhere to lock-in and costly licensing models.

But while Citrix seems to see that as an eventuality that needs to be vigilantly guarded against, Mickos says Eucalyptus is more about providing an open alternative to VMware, Microsoft, IBM, or whoever is pushing for proprietary virtualization ecosystems, providing compatibility with the dominant cloud standard (whatever it may be) while enabling portability and extensibility. It’s up to customers, Mickos says, to push back against vendors and demand openness in their cloud infrastructure.

Nebula’s Take

Kemp, formerly NASA’s first-ever CTO and a co-founder of OpenStack itself, can only speak officially for Nebula, the company he co-founded with the self-appointed mission of creating a private, enterprise-grade cloud platform that can run on commodity hardware. Regardless, Kemp is one of the foremost evangelists for OpenStack as a project and as a philosophy.

Much of my talk with Kemp touched on the same points as when I spoke with him at April’s OpenStack Design Summit. But he largely dismissed the philosophy of both Citrix and Eucalyptus. I mentioned my observation that it seemed like Mickos and Dholakia’s philosophical alignment made it seem that they were ganging up on Kemp and OpenStack on stage.

“I think that’s what happens when you’re ahead,” Kemp replied.

Bravado aside, Kemp reiterated the idea that OpenStack isn’t against a vendor, but rather a philosophy. In fact, when it comes to VMware or Microsoft, Kemp says that “OpenStack has no opinion.” OpenStack isn’t a cloud operating system or a product, but rather 17 components of a platform that a service provider or enterprise can build something unique on top of. The service provider may compete with VMware, but OpenStack is focusing on its core technology.

Kemp agreed with my suggested analogy that OpenStack is to the enterprise what the Linux kernel is to Red Hat. That’s actually part of Mickos’ skepticism around OpenStack, however – if it’s 17 loosely coupled components and not a product, then what is it?

But Kemp’s not buying it. Maybe Eucalyptus and CloudStack and, yes, Amazon Web Services have tons of adoption. But he rejects customer numbers as the measure of success. Enterprise-grade, production-ready OpenStack-based clouds may be relatively few and far between at this point, but that’s because the developers of the fledgling OpenStack Foundation are taking the time to get it right and build a core for real, true next-generation cloud platforms.

Anybody selling an OpenStack-based product (emphasis on “product,” distinct from cloud services of the type Rackspace is delivering on an OpenStack core) right now is secretly in the consulting business, Kemp said, because it’s still a ways from full and total maturity. But on the flip side, he says that when it’s really there, OpenStack is going to be able to help service providers address a much wider group of use cases.

Amazon Web Services and the public cloud will always have a place in the market, Kemp says, but with the increasing need for high-speed analytics, big data, and other applications that require high-performance, low-latency transport of huge amounts of data, Amazon Web Services will find itself on the wrong side of a geographical divide.

No easy answers

It’s hard not to get caught up in the widely-noted OpenStack optimism while talking to Kemp or Moorman. To OpenStack’s true believers, the foundation is really creating the next evolution of the cloud. And even to those a little more skeptical, it’s equally hard to argue with Moorman’s premise that competition breeds innovation and openness.

But at the same time, Mickos was equally right when he noted during the panel that young developers, especially startups, are largely going to Amazon Web Services, and that not having a native AWS compatibility layer is problematic for building a business.

It’s a tricky situation, and Kemp and OpenStack are playing the long game, betting big on the notion that the market really wants what they have cooking. In the meanwhile, a cloud API standard, open or otherwise, seems increasingly unlikely.

And in the short term, Amazon Web Services and its API are only getting more and more entrenched, which in turn reinforces CloudStack and Eucalyptus’ position. As I’ve noted before, despite the slow-but-steady approach, if OpenStack wants to vindicate the buzz and keep the momentum going, it may well need to pick up the pace.