PostgreSQL Performance on OpenBSD

From: Martin Foster <martin(at)ethereal-realms(dot)org>
To: pgsql-general(at)postgresql(dot)org
Subject: PostgreSQL Performance on OpenBSD
Date: 2003-05-15 05:22:44
Message-ID: 3EC32429.6060408@ethereal-realms.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

http://www.postgresql.org/docs/aw_pgsql_book/hw_performance/0.html

I was taking a look at the above document, and found it very
interesting. Essentially, it talks about two very important
configuration features being the Cache and Sort sizes.

Now, I was wondering how accurate it is when looking at this.
Essentially, the author recommends using 25% of the available RAM for
the cache and 2-4% for the sort buffer. Since units are available in
8K units, the numbers seem to come across as a bit high for a machine
with 512Megs, dedicated to the PostgreSQL server.

For example a conversion of Megabytes to Kilobytes then to 8K units:
128 x 1024 / 8 = 16384 (Cache)
20 x 1024 / 8 = 2560 (Sort)

My database has a lot of information that is being queried over and over
due to the fact that it's running a web based application. So for me,
a large cache makes sense as it will allow information to be queried
without the associated expensive I/O.

However are the above numbers realistic? This would require some
fairly customized kernels for OpenBSD to allow for numbers that high as
well? And what about the Write Ahead Log (WAL) would a number close to
20-64 megs seem as realistic and proper?

Martin Foster
Creator/Designer Ethereal Realms
martin(at)ethereal-realms(dot)org

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Himmet Karaman 2003-05-15 05:23:26 TESTING the DATABASE
Previous Message Tom Lane 2003-05-15 02:39:07 Re: server process segfaulting