Does it compress only the key and values, or it compresses the entire HFile? I read lzo is very efficient when you have redundant stuff, and I was thinking HFiles have a lot of redundant stuff since the whole key:value is stored for each entry. Are there two kinds of compression in HBase - at the CF level (which we specify as COMPRESSION=>'lzo' while creating the table) and at the HFile level (hfile.compression property in hbase-site.xml file as mentioned in this thread: http://www.apacheserver.net/hbase-compression-at1061265.htm). Which is better in general?
Thanks, Hari
