Re: memory requirements question

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Adam Ulmer <ulmer(at)soulgamer(dot)com>
Cc: pgsql-hackers(at)postgreSQL(dot)org
Subject: Re: memory requirements question
Date: 1999-08-24 20:36:04
Message-ID: 3729.935526964@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Adam Ulmer <ulmer(at)soulgamer(dot)com> writes:
>> can anyone tell me where I need to look so that I can calculate the
>> amount of shmem req'd by postmaster and N backends?

Easiest way is to try it and see ;-). AFAIR the space allocated per
backend is miniscule compared to the space per disk buffer, so you could
use "8K per buffer plus some constant" as a good first approximation.
If you want to know what the delta per backend is, then try a few
different -N values with fixed -B and look at what ipcs says...

> different invocations of postmaster with -N options set at 16, 32, and
> 1024 (which I gather means -B options of 32, 64, and 2048).

We require a *minimum* of 2 buffers per backend --- if you have too
few buffers then you'd lose performance due to contention for buffers.
More is probably a good idea. The ideal function is probably some fixed
number like a few dozen (to cache the system tables) plus X per backend,
where I suspect X should be more like 5 to 10. But I don't know that
anyone has really tried to measure what a good choice is for -B versus
-N. I've cc'd this to pghackers in case anyone there has results to
share.

regards, tom lane

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 1999-08-24 20:51:58 Re: [HACKERS] vacuum process size
Previous Message Tom Lane 1999-08-24 20:21:24 Re: [HACKERS] vacuum process size