The space of highly scalable key/value stores is pretty crowded. The roundups at mindstorms and metabrew should give a pretty good feel for what’s out there. Two projects I recently learned about that aren’t covered by either are Keyspace at the more Dynamo-like end of the spectrum, and Riak at the more S3-like end. Both look quite interesting, and I recommend some of the Riak pages as an introduction to things like consistent hashing and Brewer’s CAP Theorem as well. I hope I get a chance to play with them both, and if I do I’ll probably drop a note or two here.