Postgres Pro
Downloads
Home   >   mailing lists

Thread: filesystems benchmark

filesystems benchmark

From
Adrian Moisey
Date:
Hi

I'm trying to compare different filesystems for postgres using pgbench.
  The problem I've seen is that my IO wait is very very low.  Is there a
way I can get pgbench to do more to push that up a little?

--
Adrian Moisey
Acting Systems Designer | CareerJunction | Your Future Starts Here.
Web: www.careerjunction.co.za | Email: adrian@careerjunction.co.za
Phone: +27 21 818 8621 | Mobile: +27 82 858 7830 | Fax: +27 21 818 8855

Re: filesystems benchmark

From
"Vladimir Sitnikov"
Date:
I'm trying to compare different filesystems for postgres using pgbench.  The problem I've seen is that my IO wait is very very low.  Is there a way I can get pgbench to do more to push that up a little?
Why do you expect to see high IO wait?

Does your database fit in system memory? If you do not modify the data and it is not large enough, it is unlikely you will hit IO wait.

Do you have write cache enabled for the file system/device? Even if you have some DML statements, it would not be that easy to saturate write cache enabled storage (especially, when it fits in memory).


Regards,
Vladimir Sitnikov

Re: filesystems benchmark

From
Greg Smith
Date:
On Mon, 15 Dec 2008, Adrian Moisey wrote:

> I'm trying to compare different filesystems for postgres using pgbench.  The
> problem I've seen is that my IO wait is very very low.  Is there a way I can
> get pgbench to do more to push that up a little?

Increase the database scale until the wait is what you expect.  See
http://www.westnet.com/~gsmith/gregsmith/content/postgresql/pgbench-scaling.htm
for details.

The latest filesystem performance comparison I'm aware of is at
http://www.baconandtech.com/files/filesystem_io.pdf

You might also take a look at that to make sure you're not re-mapping
territory that's already been explored.

--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD