[Bioclusters] Suggestions requested on HPC.

[Lijo]

Hi,

Our lab plans to erect a HPC for molecular modeling and docking. Now we face
flood of executives from different high profile brands with their
customized efficiency.
Does the performance is free from the brands?

Expecting suggestions !

Lijo M Antony

[George Magklaras]

I can't give you specific suggestions unless you tell me a bit about the
scale of your envisaged molecular modeling and docking (size of data
set, number of cores, etc). So, in a very abstract way, my view is that
performance is often independent of the large brands and more dependent
on the local sysadmin resources and how good they work with the work
developers. This is true for small to moderate scale (500 cores say)
installation and for filesystems whose size can fit well within the
limits of DAS (directly attached storage, as opposed to fiber channel
and other storage solutions). You could find very good deals with
independent smaller scale hardware manufacturers at a fraction of the
cost of the big boys. However, if your envisaged scale exceeds the
moderate level and/or your simulation requires some sort of bespoke
interconnect, then the big boys might have more resources to deal with
annoying driver issues and help your local resources get more juice out
of your setup.

GM
--
--
George Magklaras

Senior Computer Systems Engineer/UNIX Systems Administrator
EMBnet Technical Management Board
The Biotechnology Centre of Oslo,
University of Oslo
http://folk.uio.no/georgios

Post by Lijo
Hi,
Our lab plans to erect a HPC for molecular modeling and docking. Now we face
flood of executives from different high profile brands with their
customized efficiency.
Does the performance is free from the brands?
Expecting suggestions !
Lijo M Antony

[Joe Landman]

Post by Lijo
Hi,
Our lab plans to erect a HPC for molecular modeling and docking. Now we face
flood of executives from different high profile brands with their
customized efficiency.
Does the performance is free from the brands?

Hi Lijo:

Depending upon the solution offering, performance is quite a bit more
aligned with system configuration/design issues, OS issues, and
associated. Opteron/Xeon systems from many vendors will generally have
similar performance if you isolate the computing performance. It is how
the overall system is designed that will have the most dramatic impact,
and to a lesser degree, though nearly as important, what you have in
terms of system software configuration.

We run into designs that have no hope of working reasonably well,
through massively over-engineered systems. Many of the "designs" simply
throw together brand name gear or various brands from various suppliers,
and don't focus upon what it is you are doing. The latter is the
important aspect. Which software for modeling and docking are you going
to be using, how large are your models, what are the expected run times,
how many runs do you need to accomplish per day?

If you can answer those problems, we may collectively be able to
render advice. If (like many US universities) you have pre-existing
purchase agreements with only one vendor, you will likely be quite
constrained on what you can purchase.

The more information you can share on what you are doing, the more we
can help.

Joe

Post by Lijo
Expecting suggestions !
Lijo M Antony

--
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web : http://www.scalableinformatics.com
http://jackrabbit.scalableinformatics.com
phone: [phone]
fax : [phone]
cell : [phone]

[Angulo, David]

If you're doing Molecular Dynamics (MD) or similar, the MD-Grape at Riken in Japan is orders of magnitude faster and better proce/performance. I'm not sure if they're selling them, though. Riken is similar to our national labs.

Dave Angulo
________________________________________
From: bioclusters-bounces at bioinformatics.org [bioclusters-bounces at bioinformatics.org] On Behalf Of Lijo [lijo.skb at gmail.com]
Sent: Saturday, August 02, 2008 12:00 AM
To: bioclusters at bioinformatics.org
Subject: [Bioclusters] Suggestions requested on HPC.

Hi,

Our lab plans to erect a HPC for molecular modeling and docking. Now we face
flood of executives from different high profile brands with their
customized efficiency.
Does the performance is free from the brands?

Expecting suggestions !

Lijo M Antony

--
Centre for Bioinformatics,
University of Kerala, India.
[phone](res)
_______________________________________________
Bioclusters maillist - Bioclusters at bioinformatics.org
http://www.bioinformatics.org/mailman/listinfo/bioclusters

[J.W. Bizzaro]

Post by Angulo, David
If you're doing Molecular Dynamics (MD) or similar, the MD-Grape at Riken in Japan is orders of magnitude faster and better proce/performance. I'm not sure if they're selling them, though. Riken is similar to our national labs.

This appears to be the product website:

http://www.peta.co.jp/pci-x-en.html

Cheers,
Jeff