There are quite a few free/open-source key-value and document-oriented databases available now.
CouchDB and
MongoDB are two most interesting from my perspective. They have native JavaScript/JSON support and both allow simple map-reduce jobs written in JavaScript to be sent straight to the data store for processing. You get some map-reduce with your datastore, as compared to getting a datastore (
HBase) wtih your MapReduce (Hadoop).
Both the Google and Amazon datastores are very easy to use, but Google's is only available from the
AppEngine runtime environment. Amazon's
SimpleDB is available outside of EC2, but the latency is much too high to be very useful.
Cassandra (which came from Facebook) looks solid from a design point-of-view, but its interfaces are quite low-level, creating a large barrier to adoption.
So the Web 2.0 applications of all of these products are pretty obvious. How will they be used in the enterprise?
