Eucalyptus started out in the research labs at UCSB about a year ago but the coding. It's part of an NSF funded project called V-Grads. The goal of V-Grads is to create a software infrastrucure that gives Grid and grid-like programs a uniform execution target regardless of how the resources are managed.
Every year the Eucalyptus team demos how the applications are managed across Grids and cluster and this year they were slated to demo on Amazon EC2. In that process they realized that the software would allow them to basically create their own EC2 (in addition to being able to manage EC2 itself.)
Eucalyptus is architected to treats nodes as resources and each processing task is divided into per-resource components. There is a web services component on each head node and on each cluster node. The Eucalyptus Cloud Controller interacts with the Clouds to manage the resources.
With the students flowing back to campus Wolski said they are planning a 6-8 week major re-factoring in order to solidify the internals. All interfaces will remain the same so there should be minimal impact on existing users and developers.
All of the project contributors are part of the V-grads programs and there are some random other developers that are interested--also in academia. So far there haven't been any code contributions from other parties.
With the Cloud market so new, the UCSB team hasn't yet decided what they are going to do yet in terms of just keeping it as a project or pursuing a commercial endeavor. Regardless they plan to keep developing Eucalyptus as part of their academic pursuits.
One idea that Rich and I discussed was the idea that Eucalyptus could be used to build you a "thing" that looks like whatever Cloud infrastructure you like and then you could deploy it internally. And once you are running a Eucalyptus based Cloud internally you could then manage other Cloud resources from your enterprise and decide what components cross the firewall.