VMware's annual conference was already one of the hottest shows of the year in 2009. This year, the event solidified itself as the place to be if you're in the IT industry. While developers were certainly not the focus of this show, there are an awful lot of interesting new technology and wild new ideas here that will be showing up in development environments during the coming years.
If you've been reading SD Times at all, you've no doubt noticed that I'm on a big "Data Center Operating System" kick. Before VMworld, I knew there were a lot of companies trying to build this type of software: an API and services layer that goes on top of the virtualization layer, giving developers access to the functionality they require to make cloud worth while.
But after VMworld, I now see that just about everyone in the world has a data center operating system startup. While some, like Nimbula and Eucalyptus, are already making waves with users and developers, there was another two dozen such companies at the show that I had never heard of. And even after speaking to them about their projects, I still don't remember them.
So I felt that a blog post was needed here. I wanted to clarify for you, dear readers, the situation around such DCOS companies. At present, no one can sell you a DCOS. VMware looks to be the most likely to offer such a product, and at VMworld they all but released vSphere, the management and control layer for their DCOS. But by the end of the show it was clear that this is still software in progress. Most companies aiming for this market are solving problems that don't really exist yet. Still, others are building out they types of services developers will eventually need, but their current incarnations are far too slow,unstable and untested to even be considered.
I suppose this is the real take away: Don't jump into the data center OS boat just yet. Because these systems will include services and APIs, developers will have to tie themselves into these systems upon installation. That means moving to another DCOS later on will be incredibly difficult. And that's vendor lock-in. While the DCOS will be super important to your stacks in five years, it's also a major move for your IT folks, so be careful and wait for a leader to emerge before you spend millions on more infrastructure software.
Here's my real advice: if you have to have a DCOS, make sure it's one you can test out on your desktop. Eucalyptus is a perfect example of this. While it's certainly too early to push Eucalyptus to a live private cloud for enterprise use, it's the perfect time to spin up an instance on your desktop for testing purposes. It's cheaper to run your AWS app in Eucalyptus first, then push to Amazon, than it is to just push your app to the cloud and hope it works properly.
And therein lies Eucalyptus' secret sauce: those APIs and services they offer are based on existing ones inside Amazon. Everyone else has to build their own APIs with their own behavior. That's going to be a huge advantage for them.