Hey, Data in HBase is compressed upon compaction/flushes (i.e. upon creation of the storefiles). Hence the compression is also done over blocks of data (akin to SequenceFiles) and is efficient. The memstore isn't kept compressed nor is the WAL.
RPCs in Apache HBase aren't compressed yet, but https://issues.apache.org/jira/browse/HBASE-5355 tracks it if you want to take a look. HLog (WAL) compression is also available now, via https://issues.apache.org/jira/browse/HBASE-4608 (Its in 0.94) On Wed, Apr 18, 2012 at 4:26 PM, Rita <rmorgan...@gmail.com> wrote: > How is data compressed in hbase? Is there data compressed after it hits the > disk or does the client do the compression and sent it to the region > server? It would be more efficient once it goes to the memstore and sees > what other data resembles it so compression is more effective? > > > -- > --- Get your facts first, then you can distort them as you please.-- -- Harsh J