[Bioclusters] error while running mpiblast

James Cuff jcuff at broad.mit.edu
Wed Mar 2 00:59:11 EST 2005

On Wed, 2 Mar 2005, Joe Landman wrote:

>    it is quite possible that mpiblast will scale better than NCBI blast 
> on this system.  mpi forces you to pay attention to locality of 
> reference, so you tend to do a good job partitioning your code (that is, 
> if it scales).  NCBI is built with pthreads, and I haven't seen it scale 


See - I told you that Joe knew his stuff...  (old school, with a touch of

> Lucas sent me a note indicated that in 1.3.0 they allow for shared and 
> local to coexist.  Aaron/Lucas, if you are about, could you clarify some 
> of this?  I don't want to lead people astray (and I will need to update 
> the SGE tool).

*blush* it did actually work for me with (local/local) on our cluster, but
it did keep moaning about this darn thing called 'blossom'...  if only I
knew what a 62 year old flower had to do with genome analysis ;-)

Sorry in all seriousness, a couple of weeks ago, I pushed through one of
"them there lazy, throw in a genome get out an answer, I just can't be
bothered to chunk and overlap this sucker" problems on 200 nodes.  I got
about 147 nodes worth of 'throughput', but I got the answer in really
short time, well 147 times shorter (factoring in my lazyness) to be
precise :-).

mpiblast works.  Really very well for certain problems.  There I said it.

Guy and Tim will probably never forgive me...  I think I may have been the
original 'embarrassingly parallel is the only way, nothing else will ever
give the throughput, yada, yada' advocate...

> Note: We have not built the mpiblast RPM for Itanium (nor for that 
> matter, any of our RPMs).  Is there any interest in this?  Curious.

Shame they cost so darn much, well ours do, but folk keep demanding me to
cram 64GB in them for something called whole genome assembly.  I just
can't for the life of me understand why they cost so much :-)



James Cuff, D. Phil.
Group Leader, Applied Production Systems
Broad Institute of MIT and Harvard. 320 Charles Street,
Cambridge, MA. 02141.  Tel: 617-252-1925  Fax: 617-258-0903

More information about the Bioclusters mailing list