Message-ID: <46D471BF.5040803@drivefaster.net>
Date: Tue, 28 Aug 2007 13:04:31 -0600
From: Dan Harris <fbsd@drivefaster.net>
User-Agent: Thunderbird 2.0.0.6 (Macintosh/20070728)
MIME-Version: 1.0
To: PostgreSQL Performance <pgsql-performance@postgresql.org>
Subject: Re: Performance problem with table containing a lot of
 text (blog)
References: <Pine.HPX.4.62.0708282016250.3324@purple.bdb.fi>
In-Reply-To: <Pine.HPX.4.62.0708282016250.3324@purple.bdb.fi>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Kari Lavikka wrote:
> Hello!
> 
> Some background info.. We have a blog table that contains about eight 
> million blog entries. Average length of an entry is 1200 letters. 
> Because each 8k page can accommodate only a few entries, every query 
> that involves several entries causes several random seeks to disk.  We 
> are having problems with queries like:
> 
> 1) give me a list of months when I have written someting
> 2) give me id's of entries I have written on month X year X
> 3) give me the number of blog entries my friends have written since last
>    time

I didn't see your schema, but couldn't these problems be solved by storing the 
article id, owner id, and blog date in a separate table?  It seems that if you 
don't actually need the content of the blogs, all of those questions could be 
answered by querying a very simple table with minimal I/O overhead.