C
ChUIMonster
Guest
The "c" configuration screen would tell us interesting things too. My initial impression is that you must have -B set quite small. You have quite a few OS reads with very low demand. So maybe you are scanning those tables. If that is intentional then I suppose it is ok. OTOH you seem to be updating as many records as you are reading. 3,800 TRX/sec is not horrible. But you could probably improve things. You might want to start by looking at the code and grouping multiple updates into a TRX. 50 or 100 at a time for instance rather than 1 update per TRX. You have a lot of empty bi buffer waits. I would increase -bibufs. I don't know what your bi cluster size is. You should run a dbanalys and put the output in the $PROTOP/dbanalys directory -- that way you can see how this activity compares to the table sizes and index characteristics. (ProTop should do this for you in the background if you use the "A" command.) I doubt that you need 3 APWs. After-imaging is disabled. Presumably this is a throw-away db? IO response is not very good. Is this a SAN? Or slow rotating rust? BogoMIPs is decent. That's a reasonably modern CPU.
Continue reading...
Continue reading...