Re: Good News re count(*) in 8.1

Kevin,

On 2/22/06 8:57 AM, "Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> wrote:

> I hesitate to raise this issue again, but I've noticed something which I
> thought might be worth mentioning. I've never thought the performance
> of count(*) on a table was a significant issue, but I'm prepared to say
> that -- for me, at least -- it is officially and totally a NON-issue.

Cool! Kudos to Tom for implementing the improvements in the executor to
move tuples faster through the pipeline.

We see a CPU limit (yes, another limit) of about 300MB/s now on Opteron 250
processors running on Linux. The filesystem can do 420MB/s sequential scan
in 8k pages, but Postgres count(*) on 8.1.3 can only do about 300MB/s. This
is still a very large improvement over past versions, but we'd always like
to see more...

- Luke