On Nov 25, 2007 8:06 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> The TPS numbers bounce around by 1% or so on repeated trials, so I
> wouldn't put too much faith in small differences. What it looks like
> to me is that it's all about the stats collection overhead. The drop on
> 01-17 corresponds to autovac and stats_row_level being turned on by
> default. The improvement on 03-02 is the fix for the problem that the
> stats collector process wanted to write the stats file way too often,
> and the improvement on 04-30 comes from rate-limiting stats messages
> from individual backends to the stats collector.
Using pgbench -n -S -c 10 -t 100000, I also have CVS tip as fast as
CVS from january. But using my set of queries, it's not.
It's hard to tell what is exactly causing this but the only difference
between both is that mine is based on varchars and yours is based on
integers so as Greg suggested it, maybe the culprit is the packed
varlenas patch.
I'll try to measure the overhead of this patch alone.
--
Guillaume