[Benchmarking] Some thoughts about a disk bound test
Seven (aka Arnulf)
seven at arnulf.us
Mon Sep 13 03:26:14 EDT 2010
On 09/13/2010 09:15 AM, Andrea Aime wrote:
> Hi,
> I was thinking about the possibility of making a disk bound
> test next year.
> The current server setup is quite un-balanced on that point of
> view: we have 8 cpu's (which allows for a good cpu bound testing)
> but the disk subsystem is low performance, single disk with rather
> low throughput and high seek times (mind, I'm not complaining,
> I'm very grateful we have such hardware to start with, just making
> an honest assessment of what I think would be necessary to perform
> a disk bound test).
>
> If next year we're going to make a I/O bound testing we should make
> sure we have a better disk subsystem, which also implies we should
> look for funding to get one.
>
> I believe at the very least we should look for a RAID-5, 3 disk setup,
> to be installed on both WMS servers and on the database server
> as well. Maybe three caviar black drives? However each of them
> is quite pricey...
> There is also the option of software vs harward RAID.
> I've heard good things about the software RAID built into the
> Linux kernel, no idea if Windows has anything similar, and how
> good that is.
>
> Funding options:
> - OSGeo itself (ok, I'm probably dreaming here)
> - every team taxes itself. If we're 8 and we go for a software
> raid solution this could be around 100-200$ per team?
> - we find some test sponsor that wants to stick its label
> on the presentation somewhere
Andrea,
thanks for starting these ideas right away. OSGeo's role can be to help
seek a sponsor for this and offer visibility. At the last board meeting
we have started a dialog about how we can do this and combining your
first with the last option is very viable.
...and while I am at it: Thanks again to this list and the teams behind
it for making it possible. This is what we are all about and it seems
like we have all learned quite a lot. There is more diversity and we can
see reasons each one of them to enhance in their particular domain.
Best regards,
Arnulf.
> Opinions, suggestions?
>
> Cheers
> Andrea
>
>
--
Exploring Space, Time and Mind
http://arnulf.us
More information about the Benchmarking
mailing list