We're working on an open source, high performance, distributed database modeled after Bigtable. It differs from HBase in that it is a more faithful implementation of the design outlined in the Bigtable paper and is written in C++. You can check out the project and download an "alpha" release here: http://www.hypertable.org/
It is designed to run on top of any filesystem, but HDFS has been our primary development filesystem. We did a preliminary performance test using the AOL query logs and described the experience here: http://code.google.com/p/hypertable/wiki/PerformanceTestAOLQueryLog - Doug
