Tuesday, April 8, 2008

SPOFs and Clouds

I'm really excited about the Cloud computing frenzy in the marketplace right now. So many under utilized CPU cycles within large corporate data centers (eBay, Amazon, Google, AOL, Microsoft, others) are begging for communal use. Amazon leads the pack by far right now, but I'm anxious to see how IBM plays out. They feel like the sleeping giant in all of this.

While we all benefit from near zero hosting costs in existing Cloud scenarios today, I'm concerned we're building a deep dependency on a fragile backbone. Amazon's AWS outages have tremendous ripple effects already, and exemplify my concerns.

The winning scenario is one that provides virtualization of arbitrary images (Google's current, albeit early preview, release only allows uploads of Python apps), in a highly distributed network of data centers. That's one of the reasons I think IBM wins this in the end. They'll let the higher level players like Google and Amazon vet the functionality, feature set and APIs, then swoop in with the ideal multi-data-center topography.

Service Level Agreements are going to become increasingly necessary for apps to rely on this kind of virtualization, and that's another reason why I think IBM will pull this off in the end. They have decades of experience pulling these kinds of commitments and SLAs together.

I can't justify building significant web infrastructure apps on such an immature framework. Either I need more financial pressure to drive me into the risky Cloud, or the model needs to settle into a reliable state.

No comments: