IUBio

Linux vs Unix for doing bioinformatics

Tim Cutts timc at chiark.greenend.org.uk
Wed Mar 1 12:32:18 EST 2000


In article <38BD45D0.8CE3D50A at embl-heidelberg.de>,
Steffen Schmidt  <schmidt at embl-heidelberg.de> wrote:
>
>We are running BLAST searches on Linux based PentiumIII-Servers - we
>have severe problems (i.e. the server crashes) when we use e.g. blastpgp
>and a database which is accessed via NFS. This is a known bug to Linux
>as far as I know (If someone knows better please answer). So the best
>way to use BLAST is by using locally stored databases and everything is
>fine!

Running BLAST on NFS-mounted directories is a bad idea; BLAST 2 memory
maps the file, and the performance of a mmap'ed file across NFS is
appalling.  Much better to maintain a cache of most recently used
databases on the machine's local disk, and blast that.  In other words,
you have a shell script that does:

if (! have_requested_database)
  while (! have_room_for_requested_database)
    remove least recently used local database
  endwhile
  fetch requested database from it's remote source (e.g. NFS or FTP
  server)
endif
run_blast on local copy

This removes load on your network (absolutely vital if you're using a
farm) and makes the BLAST searches run much more quickly.

As you say, Linux' NFS performance is poor even by NFS' rather poor
standards...

Tim.






More information about the Bio-soft mailing list

Send comments to us at biosci-help [At] net.bio.net