[Bioclusters] Optimal database fragments in mpiBLAST

Joseph Landman bioclusters@bioinformatics.org
Fri, 14 Nov 2003 10:54:13 -0500


Hi Malay:

   Remember that the hyperthreading gives virtual CPUs.  HT does help a 
bit on blast, though mpiBLAST is using 1 thread per MPI thread.  I would 
advise against using more than 1 MPI thread per CPU, simply on the basis 
of resource contention.  Each thread is going to consume 1) memory, 2) 
memory bandwidth, 3) CPU time.  The HT will not help more than 20-30% on 
blastall, and you may lose all you have gained by forcing a smaller 
query size (hence more frequent network traffic).

   That said, I do know some groups using HT, and a ratio of 2 for 
number of threads to number of CPUs.  Wall clock times for the results 
are not any better (actually worse due to the resource issues indicated).

   Best thing to do is to make sure the database indices fit into RAM. 
  Leaving extra room is a good idea.  Too much network traffic could 
hurt your server's pipe, so balance the size against the traffic.

Joe

Malay Kumar Basu wrote:

> Hello Gurus:
> 
> Here is my cluster config:
> 
> Master - cpu 2 Xeon with hyperthreading 2 GB RAM
> 4 x nodes each - cpu 1 P4 hyperthreading 1 GB RAM
> 
> SGE recognizes total 12 cpus.
> 
> When hyperthreading on the whole setup can have 12 cpus, otherwise 6. 
> What should be the optimal BLAST database fragments for mpiBLAST?
> 
> Malay
> 
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters

-- 

Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615