Date: Tue, 22 May 2007 12:10:03 +0200
To: "Gregory Stark" <stark@enterprisedb.com>
Subject: Re: Postgres Benchmark Results
From: PFC <lists@peufeu.com>
Cc: pgsql-performance@postgresql.org
Content-Type: text/plain; format=flowed; delsp=yes; charset=utf-8
MIME-Version: 1.0
References: <op.tsmrv71ecigqcu@apollo13> <87hcq5l1c7.fsf@oxford.xeocode.com>
Content-Transfer-Encoding: 7bit
Message-ID: <op.tsp3u1becigqcu@apollo13>
In-Reply-To: <87hcq5l1c7.fsf@oxford.xeocode.com>
User-Agent: Opera Mail/9.10 (Linux)

> Note that while the average hits/s between 100 and 500 is over 600 tps  
> for
> Postgres there is a consistent smattering of plot points spread all the  
> way
> down to 200 tps, well below the 400-500 tps that MySQL is getting.

	Yes, these are due to checkpointing, mostly.
	Also, note that a real forum would not insert 100 posts/s, so it would  
not feel this effect. But in order to finish the benchmark in a correct  
amount of time, we have to push on the inserts.

> Some of those are undoubtedly caused by things like checkpoints and  
> vacuum
> runs. Hopefully the improvements that are already in the pipeline will  
> reduce
> them.

	I am re-running it with other tuning, notably cost-based vacuum delay and  
less frequent checkpoints, and it is a *lot* smoother.
	These take a full night to run, so I'll post more results when I have  
usefull stuff to show.
	This has proven to be a very interesting trip to benchmarkland...