[Bioclusters] error while running mpiblast

Aaron Darling darling at cs.wisc.edu
Thu Mar 3 18:30:34 EST 2005

Tim Cutts wrote:

> On 2 Mar 2005, at 5:59 am, James Cuff wrote:
>> 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...
> Aargh - he's gone over to the Dark Side!!!
Haha!  You guys crack me up :-)   You will be assimilated.  Resistance 
is futile.

> Seriously, I agree with you.  MPIBlast gets you fast turnaround for 
> single very large searches.  I still think for the things Sanger are 
> doing we do better with the embarrassingly parallel model, but I 
> wouldn't claim that it's always the right solution (at least not any 
> more, he said, covering his tracks in case he's ever said exactly that 
> somewhere in the past)

For whatever it's worth, my opinion is that there are far better -- 
faster *and* more sensitive -- local alignment algorithms than BLAST and 
it's a shame they haven't come into wide use.  If (when) NCBI takes one 
of those better algorithms and calls it BLAST I bet it will get used.  
People know NCBI BLAST, people trust NCBI BLAST.  As long as that 
remains true there will be a place for parallel NCBI BLAST, and with 
huge databases even database segmentation will have a place.  What I 
really look forward to is for somebody to come up with a clever 
compression and searchable indexing scheme that accounts for the huge 
amount of redundancy in big databases like nt.  Then we won't need 
mpiBLAST anymore.


More information about the Bioclusters mailing list