[BioBrew Users] BioBrew mpiBlast tutorial 1.3 time?

[Raymond Chan]

Hi Glen or anyone who knows,

 

I am testing ROCKS 3.3.0 w/ Biobrew roll on AMD Dual Opterons w/ 2 gb ram.
I've got a head node distributing a BLAST using mpiBLAST to two compute
nodes hooked together in a 10/100 switch (yes slow switch, but I'm just
testing things out until my new equipment comes in).

 

I don't know if you have all seen it, but Glen has a tutorial to run
mpiBLAST using the Hs.seq.uniq collection w/ a sequence file called IL2RA.
I was wondering if any of you have followed the tutorial and ran a
mpiformatdb then a mpiblast, and if you knew how long it took (I know
everyone's setup is different in hardware, # nodes, etc, so time will be
different for everyone, but I want a vague idea of if this was a "slow" or
"fast" process?)

 

I'm asking because I ran mpiblast 1.3 (which was originally installed w/
BioBrew), and this particular search took about 43 seconds.  When I upgraded
to mpiblast 1.4, the search took an outrageously greater amount of time
(18minutes, 23s).  I wonder what the discrepancy is that caused this, as my
setup has not changed other than upgrading to 1.4.  I am also not sure if a
search of this kind w/ this particular sequence/DB combo is supposed to
actually take this long on my equipment?  For all I know, the 43 seconds in
my original setup w/ v1.3 could have been wrong and too fast, and the time
it took for v1.4 is more realistic in my hardware configuration?  Knowing
the general speed you have encountered allows me to investigate further into
if v1.4 is indeed the problem, and then I can ask the mpiBLAST user group.

 

Any ideas?


Thank you,

Ray

Univ of CA-Davis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://bioinformatics.org/pipermail/biobrew-users/attachments/20050718/96199ae0/attachment.html