From: | Alfred Perlstein <bright(at)wintelcom(dot)net> |
---|---|
To: | Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> |
Cc: | Matthew <matt(at)ctlno(dot)com>, "'paitoon butri'" <pbt_10400(at)yahoo(dot)com>, pgsql-admin(at)postgresql(dot)org |
Subject: | Re: PostgreSQL + SMP |
Date: | 2000-12-10 20:29:23 |
Message-ID: | 20001210122923.L16205@fw.wintelcom.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-admin |
* Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> [001210 12:28] wrote:
> > * Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> [001210 12:19] wrote:
> > > > Postgre is not threaded, but every connection gets it's own process. The OS
> > > > will distribute the processes across the processors. Not sure I said that
> > > > very will.
> > > >
> > > > Basically a single connection will not be any faster with SMP, but multiple
> > > > connections will be.
> > >
> > > Actually, even a single process will be faster because all other system
> > > process will run on other cpu's.
> >
> > Depending on how idle you are, the large overhead of bus locking
> > can cause a lot of perf degredation(sp?) for the idle SMP case.
>
> True. I was thinking more of the backend having to give up the cpu to
> allow other processes to run. With two cpu's, I believe one backend can
> camp out on a cpu forever as long as the other cpu doesn't get busy.
That's true, as long as there's no contention for the locks then it's
pretty cheap.
--
-Alfred Perlstein - [bright(at)wintelcom(dot)net|alfred(at)freebsd(dot)org]
"I have the heart of a child; I keep it in a jar on my desk."
From | Date | Subject | |
---|---|---|---|
Next Message | Michael B. Babakov | 2000-12-10 20:35:46 | users work now with PostgreSQL |
Previous Message | Bruce Momjian | 2000-12-10 20:28:19 | Re: PostgreSQL + SMP |