Re: Super Optimizing Postgres

From: Alex Pilosov <alex(at)pilosoft(dot)com>
To: mlw <markw(at)mohawksoft(dot)com>
Cc: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Super Optimizing Postgres
Date: 2001-11-16 16:26:11
Message-ID: Pine.BSO.4.10.10111161125050.31668-100000@spider.pilosoft.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Sure, that'd be useful. Especially since we all can discuss what things
work, what doesn't work.

On Fri, 16 Nov 2001, mlw wrote:

> It is sort of discouraging that Postgres' default configuration is so
> conservative. A tweek here and there can make a big difference. It seems to me
> that the default postgresql.conf should not be used for a dedicated server. In
> fact, it can't because TCP/IP is disabled.
>
> In my projects I have done the default stuff, increase buffers, sort
> memory, and so on, however, some of the tunable parameters seem a bit
> arcane and are not completely clear what they do or the effect they
> may have. (some have no noticable effect, eventhough it looks as if
> they should.) I think most users, particularly those new to SQL
> databases in general, would find it difficult to tune Postgres.
>
> Does anyone think it is a good idea, to make a postgresql.conf
> cookbook sort of thing? Gather a number of tuned config files,
> annotated as to why the settings are set the way they are, and the
> machine on which they run.
>
> Particularly, I'd like to see if someone has been able to really
> understand and manipulate the planner COST options successfully.
>
> Alternatively, it should be possible to write a program that analyzes
> a target system, asks questions like: "Is this a dedicated server?"
> "How much ram do you have?" "On which volume will the database be
> installed?" Then perform some tests that mimic the cost values, and
> create a new postgresql.conf with the options tuned.
>
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo(at)postgresql(dot)org
>
>

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Stephan Szabo 2001-11-16 16:28:59 Re: Fwd: Btree doesn't work with ">" condition
Previous Message Tom Lane 2001-11-16 16:12:28 Re: import/export of large objects on server-side