Michelleberard
Technical User
Hi, I'm an Oracle DBA and we have a single, 170GB database on an AIX 5L server with 8GB RAM. The data is stored on an EMC Symm and we're using JFS2 filesystems.
There are only a handful of users on this powerful server accessing the database (which will convert to production very soon). I have optimized the init.ora parameters; Oracle's SGA is roughly 3GB. I have 12 years of Oracle DBA experience and feel almost certain it's not the database (I know, you've heard that before).
I'm suspicious of the AIX parameters, specifically minperm, maxperm, maxclient, minfree, maxfree - most of these settings are at default - and I understand that Oracle does not perform well with 20/80 minperm/maxperm. I tried tweeking it once to 10/40 and the users reported a slight improvement but some transactions were just as slow as ever. Our Sys Admin wanted these parameters changed back t their defaults. We're using asynch/io. But we're not mounting our database filesystems with direct or concurrent i/o - I'm not sure if it even matters in a SAN environment.
Our SysAdmin just applied the 5.2.0.5 patch today and we're hoping that may solve the problem but it seems a little too optimistic. Before the latest patch was applied, lsps -a showed up to 20% swap space used. It's only 3% now but the server has only been up about 10 hours.
AIX folks with Oracle experience, I'm open to any and all pointers. We have to find a solution soon!!!
There are only a handful of users on this powerful server accessing the database (which will convert to production very soon). I have optimized the init.ora parameters; Oracle's SGA is roughly 3GB. I have 12 years of Oracle DBA experience and feel almost certain it's not the database (I know, you've heard that before).
I'm suspicious of the AIX parameters, specifically minperm, maxperm, maxclient, minfree, maxfree - most of these settings are at default - and I understand that Oracle does not perform well with 20/80 minperm/maxperm. I tried tweeking it once to 10/40 and the users reported a slight improvement but some transactions were just as slow as ever. Our Sys Admin wanted these parameters changed back t their defaults. We're using asynch/io. But we're not mounting our database filesystems with direct or concurrent i/o - I'm not sure if it even matters in a SAN environment.
Our SysAdmin just applied the 5.2.0.5 patch today and we're hoping that may solve the problem but it seems a little too optimistic. Before the latest patch was applied, lsps -a showed up to 20% swap space used. It's only 3% now but the server has only been up about 10 hours.
AIX folks with Oracle experience, I'm open to any and all pointers. We have to find a solution soon!!!