Eucalyptus and OpenStack: A Discussion with Marten Mickos


In response to my recent post “Ubuntu, OpenStack, Eucalyptus: When Open-Source Competes with Open-Source”, Marten Mickos, CEO of Eucalyptus Systems (and previously of MySQL) offered to discuss the Eucalyptus strategy with me in the “hope … that we would NOT confuse you.” I took him up on it, and a lively and rather interesting discussion ensued.

This post is a summary of the major points of the discussion. (See also Marten’s presentation at Gluecon this past week.)

Happy reading!

Johannes:

would love to be less confused … The thing I don’t quite understand is Eucalyptus’ relationship to the developing ecosystem.

OpenStack now has this gigantic sucking sound and it attracts lots and lots of rather smart people from all over the place … there’s a big melting pot of rather interesting ideas going on that will likely end up producing something much broader and more interesting than originally conceived … or that could have been conceived or funded by any one of the participants.

Why not take the high road, fully support OpenStack … borrow and integrate whatever they come up with opportunistically to become API compatible with BOTH Amazon and OpenStack, and thus leverage both ecosystems?

Marten:

I agree with you that OpenStack has gathered an incredible following in a short time. Kudos to them for that. “Big melting pot of rather interesting ideas” is a great description of it. Those ideas will surely have a positive impact on the IaaS layer of the clouds.

To the extent OpenStack attracts smart people who produce superior code, we will have no hesitation in using such pieces of the software. It’s wonderful that OpenStack is driving these initiatives. In our estimation, however, OpenStack may still be two years away from serious production use.

In the meantime, we have a massive market that we are serving. Over 25,000 Eucalyptus clouds have been started up all over the world. Our product has been ready for production for over a year. We are busy serving those users (Plinga, USDA, Puma, InterContinental Hotels, Aerospace Corporation, NSA, etc.). Our business is growing faster than we can handle. These customers are looking for performance, scalability, low latency and robustness of code. We have what we ourselves consider a very attractive roadmap for our product with sexy new features, but still our users value us mostly for having a hardened, fully QA’d platform.

What drives our business is, apart from the product being robust and ready, the fact that we are fully AWS API compliant, that we focus on the enterprise market, and that we are the best alternative to VMware’s vCloud Director.

…The AWS compatibility is just a feature on the surface. Our main engine is possibly the most innovative cloud platform the world has seen.

Rackspace established OpenStack … to serve the needs of service providers. That’s an amazing market in itself, but it is not really the one we have chosen for ourselves. That’s why we may not be seen as fully supporting it. Our own product is designed to support multiple APIs, so adding OpenStack API is a straightforward thing. Given our enterprise focus, I would however estimate that we would first consider the VMware and MSFT Azure APIs. But who knows how the markets will develop, so we will maintain fully flexibility and readiness to support whatever our markets are best served by.

Does this make sense to you?

Johannes:

I’d be 100% with you (paraphrased as “traction beats lofty talk” ;-)) if we assumed that the market is IaaS — something that’s relatively well understood and with clear boundaries. But I think we’re well into what I called the “growing towers” model in my post here and PaaS is already beginning to be all about ecosystem — one of the reasons behind CloudFoundry I think. There will only be so many viable competing ecosystems — as it historically has always been around new platforms — and OpenStack is producing this rather gigantic sucking sound already (witness Ubuntu).

Marten:

Here I am not sure I agree. Human beings overestimate the short-term effects and underestimate the long-term effects. In my observation, IaaS is well understood as a concept and as a required and newly defined layer. But creating it and making it work flawlessly will take years. Trying to expand it too soon – to tie it too soon to PaaS and other layers may cause IaaS to not mature well enough or soon enough.

I could of course be wrong, but that’s my conviction. That’s why Eucalyptus is so singularly focused on what it does.

I could be wrong, but I compare the new cloud industry to what was the new RDBMS industry some 30 years ago. Back then, there were a handful of players who grew to a billion or so in revenues: Ingres, Sybase, Informix, DB2 and Oracle. Only when those products/vendors were big did the landscape start to zero in on just a few designs (of which 3 today survive).

So I am thinking that it is far too early to think that we must all get behind just one or two initiatives. What looks great today may look bad tomorrow. What causes a gigantic sucking sound today may be in disarray tomorrow. What looks bad today, or isn’t visible at all, may be the winner of tomorrow. There are really smart software developers in Russia, China, India and Brazil who haven’t come out with their IaaS platform yet. There could very well be a new team coming out that beats all of us to the goal.

Johannes:

From my point of view as an entrepreneur, I got to deal with whatever Amazon does because they dominate public cloud, with whatever VMWare does because of their footprint in enterprises, and if I can deal with one more thing that’s got to be the initiative that gives me a chance of dealing with as many others as possible at the same time, which now would be OpenStack. My impression is that many other entrepreneurs and established companies think the same way … which is the line of thinking that lead to my original comment on my blog.

If it’s really so easy to put a new API on your code as you say, then I really don’t see the downside of putting half a person on writing a press release and going to OpenStack meetings ;-)

Marten:

Sure, I fully understand your viewpoint and preference, and the one of other entrepreneurs. I can see you having a great future in the OpenStack ecosystem.

What I see differently, I guess, is the nature of the overall landscape. I believe in a huge cloud business for those who are Amazon-aligned and Vmware-competitive. That’s what Eucalyptus does.

What OpenStack stands for is a world which is Amazon-competitive and Vmware-competitive. Such an initiative is needed for keeping the industry in balance. It could grow fast. But it could also grow slowly or be limited in terms of its upside.

Thanks for great arguments, and an interesting discussion! I hope I answered your main points.


Now that I understand it better (Thanks, Marten!), this strategy does make a lot of sense. After all, Oracle’s big break was implementing IBM’s SQL better than anybody else: that seems to be the Eucalyptus plan, just this time with the Amazon APIs. The big Eucalyptus bet is that the market circumstances are indeed similar, and that OpenStack won’t be able to establish a “rival SQL” with substantial market share in the enterprise market. A bet worth making?


One response to “Eucalyptus and OpenStack: A Discussion with Marten Mickos”