[Bioclusters] Clustering, compute farming & distributed computing in

[Guy Coates]

>
>
> If its not too much work, I'd like to bug you for these Joe. Additionally,
> does anyone have any experience with Luster or any other parallel
> filesystems? Is it possible to run GPFS on non-IBM systems -- are they
> selling it?

Yes. You can run GPFS on non IBM kit, and they do sell it. As you are an
.edu you can get if for no cost through the IBM scholars program. We've
had very good experiences with it and use it heavily on our cluster for
blasting etc.

Lustre is a bit more complicated; Cluster Filesystem Inc (CFS) develop the
code, and release it under a Aladdin ghostscript type model. Release X is
proprierty, but release X-1 gets released under the GPL.

The GPL lustre code is not production ready. The current CFS version is
allegedly much better, but I've not used it. There are also 3rd party
vendors who are selling more recent CFS code bundled storage as cluster
filesystem appliances (such as HP and Linux Networx).

The only downside with cluster filesystems is that they do add some
sysadmin overhead. It probably isn't worth the hassle on small clusters
(though they are nifty things to play with if you have the time!) but if
you are running into significant IO problems, give it a go.

If you are thinking about it seriously, I can give you some pointers on
our setup.

Cheers,

Guy






> -Lucas
>
> On Wednesday, November 24, 2004 at 09:50 -0500, Joe Landman wrote:
> >
> > Third, there are a number of kernel tunables that can improve the disk
> > IO performance.  If you bug me, I can find a link for you.
> >
> >
> > You can also look at alternative architecture disks, or server mods.  If
>
> > you contact me offline, I can give you some ideas.
> >
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
>
>
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://bioinformatics.org/pipermail/bioclusters/attachments/20041124/0c908aad/attachment-0001.htm
>
> ------------------------------
>
> Message: 5
> Date: Wed, 24 Nov 2004 15:06:15 -0500
> From: Joe Landman <landman at scalableinformatics.com>
> Subject: Re: [Bioclusters] NFS performance with multiple clients.
> To: "Clustering,	compute farming & distributed computing in life
> 	science informatics"	<bioclusters at bioinformatics.org>
> Message-ID: <41A4E9B7.2060903 at scalableinformatics.com>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> I cannot say enough good things about Panasas.  I took the iobench code
> that is part of a common HPC benchmark, rewrote a portion of it to use
> MPI, and did some tests.  It is nice to see 16 compute nodes sustaining
> 1.2 GB/s write access to a single file system.
>
> It is not inexpensive, but if you need very high very scalable
> performance, it is a great design.  Each rack shelf adds file system
> bandwidth.
>
> If you want to know more about this, let me know.
>
> Joe
>
>
> jason.calvert at pharma.novartis.com wrote:
>
> >
> > You should look at Panasas, It is an appliance approach to a high
> > performance filesystem which is similar to luster.  I think Garth,
> > (founder of panasas and co-author of the RAID papers) was involved
> > with lustre.
> >
> > Jason
> >
> >
> >
> > 	*Lucas Carey <lcarey at odd.bio.sunysb.edu>*
> > Sent by: bioclusters-bounces at bioinformatics.org
> >
> > 11/24/2004 01:51 PM
> > Please respond to "Clustering,  compute farming & distributed
> > computing in life science informatics"
> >
> >
> >         To:        "Clustering,  compute farming & distributed
> > computing in life science informatics" <bioclusters at bioinformatics.org>
> >         cc:        (bcc: Jason Calvert/PH/Novartis)
> >         Subject:        Re: [Bioclusters] NFS performance with
> > multiple clients.
> >
> >
> >
> >
> > If its not too much work, I'd like to bug you for these Joe.
> > Additionally, does anyone have any experience with Luster or any other
> > parallel filesystems? Is it possible to run GPFS on non-IBM systems --
> > are they selling it?
> >
> > -Lucas
> >
> > On Wednesday, November 24, 2004 at 09:50 -0500, Joe Landman wrote:
> > >
> > > Third, there are a number of kernel tunables that can improve the disk
> > > IO performance.  If you bug me, I can find a link for you.
> > >
> > >
> > > You can also look at alternative architecture disks, or server mods.
> >  If
> > > you contact me offline, I can give you some ideas.
> > >
> > _______________________________________________
> > Bioclusters maillist  -  Bioclusters at bioinformatics.org
> > https://bioinformatics.org/mailman/listinfo/bioclusters
> >
> >
> >------------------------------------------------------------------------
> >
> >_______________________________________________
> >Bioclusters maillist  -  Bioclusters at bioinformatics.org
> >https://bioinformatics.org/mailman/listinfo/bioclusters
> >
> >
>
>

-- 
Dr. Guy Coates,  Informatics System Group
The Wellcome Trust Sanger Institute, Hinxton, Cambridge, CB10 1SA, UK
Tel: +44 (0)1223 834244 ex 7199