Christo- Do you have some reproducible test cases with performance metrics that demonstrate that these hidden parameters do indeed better performance and quantify any gains? I guess it seems very odd to me that if the default values of these hidden parameter are suboptimal then why doesn't Oracle just change them to be optimal? (I know this seems to be your point, however mine is, there must be more to it) On Wed, Sep 16, 2009 at 11:18 AM, Christo Kutrovsky <kutrovsky.oracle@xxxxxxxxx> wrote: > Short story (assuming 64 bit OS and Oracle): > > Increase limit to 4 GB (8 GB per process). Note, first parameter is in > bytes, second in KiB. > *._pga_max_size=8000000000 > *._smm_max_size=5000000 > > Even when it does spill to disk, increase IO sizes so that it's more > efficient: > *._smm_auto_max_io_size=2048 > *._smm_auto_min_io_size=256 > > The IO setting can increase a spilled sort/hash join by a factor of > 2-3, but is very SAN Cache/Usage specific. > > It still puzzles me why Oracle has limited global bound to 1GB... -- Regards, Greg Rahn http://structureddata.org -- //www.freelists.org/webpage/oracle-l