Re: Performance problems testing with Spamassassin 3.1.0 Bayes module.

From: "Merlin Moncure" <merlin(dot)moncure(at)rcsonline(dot)com>
To: "Matthew Schumacher" <matt(dot)s(at)aptalaska(dot)net>
Cc: <pgsql-performance(at)postgresql(dot)org>
Subject: Re: Performance problems testing with Spamassassin 3.1.0 Bayes module.
Date: 2005-07-28 12:20:35
Message-ID: 6EE64EF3AB31D5448D0007DD34EEB3417DCF21@Herge.rcsinc.local
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

> I'm not sure how much this has been discussed on the list, but wasn't
> able to find anything relevant in the archives.
>
> The new Spamassassin is due out pretty soon. They are currently
testing
> 3.1.0pre4. One of the things I hope to get out of this release is
bayes
> word stats moved to a real RDBMS. They have separated the mysql
> BayesStore module from the PgSQL one so now postgres can use it's own
> queries.
>
> I loaded all of this stuff up on a test server and am finding that the
> bayes put performance is really not good enough for any real amount of
> mail load.
>
> The performance problems seems to be when the bayes module is
> inserting/updating. This is now handled by the token_put procedure.

1. you need high performance client side timing (sub 1 millisecond). on
win32 use QueryPerformanceCounter

2. one by one, convert queries inside your routine into dynamic
versions. That is, use execute 'query string'

3. Identify the problem. Something somewhere is not using the index.
Because of the way the planner works you have to do this sometimes.

Merlin

Browse pgsql-performance by date

  From Date Subject
Next Message Merlin Moncure 2005-07-28 13:02:18 Re: Finding bottleneck
Previous Message Claus Guttesen 2005-07-28 11:52:03 Re: Finding bottleneck