Just keep in mind that between the two, Cassandra has improved a lot more than HBase has.
HBase can be an easy choice if you already have a Hadoop cluster and want to roll the results of Map-Reduce jobs into HBase keys.
The DataStax crew has a decent stack for turning batch/OLAP jobs into queryable keys.
If you have no need of that, want tunable consistency, favor write availability over read performance, then Cassandra might be a fit.
Just uh, don't pretend Cassandra clusters are necessarily trivial to manage just because they're homogenous.
IMHO: put off moving to any of these technologies as long as possible.
Just keep in mind that between the two, Cassandra has improved a lot more than HBase has.
HBase can be an easy choice if you already have a Hadoop cluster and want to roll the results of Map-Reduce jobs into HBase keys.
The DataStax crew has a decent stack for turning batch/OLAP jobs into queryable keys.
If you have no need of that, want tunable consistency, favor write availability over read performance, then Cassandra might be a fit.
Just uh, don't pretend Cassandra clusters are necessarily trivial to manage just because they're homogenous.
IMHO: put off moving to any of these technologies as long as possible.