Re: excessive number of db writer
- From: "Anand Rao" <panandrao@xxxxxxxxx>
- To: kevinc@xxxxxxxxxxxxx
- Date: Thu, 7 Sep 2006 12:57:22 +0530
Database writers don't "manage buffers". They get posted to work, they
grab n LRU latch....
yes, certainly. i know dbwr doesn't do buffer cache management. i was
talking in terms of the "management" of the writes. should have mentioned
that clearly...i thought it was easier for the reader(s) to understand the
complete process simply as management.
apologies for the misleading statement and thanks for pointing it out.
Hmmm...side effects. i haven't actually heard anyone complain that their
system is affected in any way (performance, reliability, response time, what
else??) when a high no. of DBWRs are configured. well, you have more no. of
unix processes, a few more semaphores....
there isn't any documented evidence that i have come across which throws
light on the subject of side effects.
it makes sense to have multiple writers when u have true kernelised asynch
io and an SMP box hosting an application with some really heavy writes.
regards
anand
--
All I need to make a comedy is a park, a policeman and a pretty girl -
Charlie Chaplin
Other related posts:
