When I run multiple TPC-H queries (DBT3) on postgresql, I found the system
is not scalable. My machine has 8GB memory, and 4 Xeon Dual Core processor
( 8 cores in total). OS kernel is linux 2.6.9. Postgresql is 7.3.18. I
q2 queries simultaneously. The results are:
1 process takes 0.65 second to finish.
2 processes take 1.07 seconds.
4 processes take 4.93 seconds.
8 processes take 16.95 seconds.
For 4-process case and 8-process case, queries takes even more time than
they are executed serially one after another. Because the system has 8GB
memory, which is much bigger than the DB size(SF=1), and I warmed the cache
before I run the test, I do not think the problem was caused by disk I/O.
I think it might be caused by some contentions. But I do not know postgresql
much. May anybody give me some clue to find the reasons?
pgsql-performance by date
|Next:||From: Dave Dutcher||Date: 2007-03-30 20:56:17|
|Subject: Re: scalablility problem|
|Previous:||From: Erik Jones||Date: 2007-03-30 16:19:09|
|Subject: Re: Shared buffers, db transactions commited, and write IO on Solaris|