Hi,
We have very bad performance issues in one of our customer PostgreSQL servers and we would like some suggestions to improve the performance for bulky updates.
When one of the biggest tables has all lines updated for example, it takes at about 30 minutes for processing. If we drop all indexes (21) and let only the primary index the same update takes 2 minutes.
As far as I could see there is no cpu nor memory bottleneck but sar is indicating 50% of iowait during those update processings.
What I'm suggesting in priority order:
a) add another 2 SCSI disks and move index data to the new drives
b) add a SCSCI controllers with RAID support and move existing disks to the new controller
c) add another disk and move pg_xlog to there
Is there any other suggestion to improve the performance of those updates?
Thank you in advance!
Carlos