[Bioclusters] MPI clustalw

Michael Cariaso bioclusters@bioinformatics.org
Fri, 07 May 2004 14:15:52 -0400

Ognen Duzlevski wrote:

>>Date: Thu, 6 May 2004 21:30:49 -0400
>>From: James Cuff <jcuff@broad.mit.edu>
>>Reply-To: bioclusters@bioinformatics.org
>>To: bioclusters@bioinformatics.org
>>Subject: Re: [Bioclusters] MPI clustalw
>>On May 7, 2004, at 11:36 AM, Michael Cariaso wrote:
>>>After seeing this abstract
>>>  http://bioinformatics.oupjournals.org/cgi/content/abstract/20/7/1193
>>>about a newly released MPI version of clustalw, coming out of of
>>>University of Western Australia
>>>I was wondering of anyone had any experience to compare it to this
>>>coming from singapore?
>>So here's my two pence on this.
>>I never really got the whole benefit of mpi/pvm speed up deal for
>>compute farms.  There I said it, I'm sorry.  :-)
>As someone who wrote a pthreads version of clustalw 1.82
>(http://bioinfo.pbi.nrc.ca/clustalw-smp) and then went on to try and do
>the same by way of pvm, I have to agree with you. I didn't see it then and
>I don't see it now :)

I think your observations speaks more to the nature of the problems that 
you've been solving, than any comment on pvm  or mpi. I've got a couple 
of tasks that involve running blast and clustalw thousands of times. But 
the output of each run forms the input for the next round. Its not 
embassingly parallel, and can't just be fed to the farm. For these 
problems the MPI versions actually make it possible for me to use the 
machines very effectively. If you tasks don't need the MPI layer, avoid 
the overhead and fire away. But these libraries are extremely valuable 
for opening up avenues that couldn't be explored otherwise.

>Digital Biology Laboratory
>University of Missouri-Columbia
>Bioclusters maillist  -  Bioclusters@bioinformatics.org