[Bioclusters] Web interface (PISE) on top of Sun Grid Engine on Mac Xserve cluster.

David Adelson bioclusters@bioinformatics.org
Thu, 30 Jan 2003 12:14:12 -0600


Greetings biocluster readers,

Forgive me if this is not the sort of message you normally get, but I wanted
to make you aware of a new cluster solution that we at Texas A&M have just
implemented for bioinformatics users.

We have a five node (10 cpu) Xserve cluster, one node serves as the head
node only, and four are workers.  Users access the system over the web and
submit jobs via PISE (Letondal C. A Web interface generator for molecular
biology programs in Unix. Bioinformatics. 2001 Jan;17(1):73-82. PMID:
11222264 ).  Pise then sends the job to a Sun Grid Engine queue, where it is
split and sent to the workers. This is the first implementation/port of SGE
for MacOS X that I am aware of. Upon completion, e-mail is sent to the user
with either a mime attachment of results or a URL to browse results.
Multiple users are thus allowed concurrent access to resources using SGE.
Command line use is also supported.

Applications at this stage include all of the biotools ported to Mac OS X by
the BioTeam ( http://bioteam.net), such as Altivec optimized BLAST, HMMER,
FASTA and Clustal and the EMBOSS suite.

Porting, configuration and installation of the cluster software was done by
The BioTeam (http://bioteam.net) for us.  As far as I know, this is the
first cluster implementation of it's kind, using PISE, SGE queue management
and OS X.  

As an example, of performance, I ran a set of 3600 ESTs against Swissprot
(BLASTX) using all 8 cpus and the job ran in about 45 minutes.  The same
ESTs against build 30 of the human genome (Database: /BlastDB/H_sapiens 1388
sequences; 2,825,300,773 total letters) took 2 hours and 10 minutes on all 8
cpus.  I have not run rigorous comparative tests yet between this cluster
and our other BLAST server (SGI Origin 3800 multiprocessor/shared memory
machine), but my gut feeling is that the Xserve cluster is running
significantly faster.

You can browse our site at:  http://xblast.tamu.edu

I have to commend The BioTeam (in particular Bill VanEtten) for their work.

Cheers,

Dave 


-- 
David L. Adelson, Ph.D.
Associate Professor - Animal Genomics
Dept. of Animal Science
Dept. of Veterinary Anatomy and Public Health
Faculty of Genetics

Postal address:
Animal Science Dept.
Texas A&M University
TAMU - 2471
College Station,TX 77843-2471
U.S.A.

Tel. +1 979 845 2616
Fax. +1 979 845 6970

http://animalscience.tamu.edu/ansc/breeding.html