These are much larger changes in results then I experienced with table compression with no indexes. The compression ratio measured in blocks was also 50%. But the run time difference was minimal. At the time I my thought was that logical IO was slower on compressed tables because of the symbol table lookups for compressed columns and rebuilding each row. -----Original Message----- From: MacGregor, Ian A. [mailto:ian@xxxxxxxxxxxxxxxxx] Sent: Tuesday, January 25, 2005 12:10 PM To: oracle-l@xxxxxxxxxxxxx Subject: Thumbs Up on Compression I built a non-partitioned, uncompressed table with 709,652,582 rows and = indexed it with a non-compressed index on (timestamp, pv_id) with the = most restrictive column first. A sample query SELECT value, timestamp, nanosecs, stat, sevr, ostat from chanarch_pepii.new_archive_data_f WHERE pv_id =3D :CUR_PV_ID AND timestamp BETWEEN :START_TIME_ORACLE_DATE AND :END_TIME_ORACLE_DATE AND ostat <> 1 ORDER BY timestamp, nanosecs -- //www.freelists.org/webpage/oracle-l