[Bioclusters] Local blast server,beowulf vs mosix + BIOITWORLDconference

[Eric Engelhard]

Jeff Layton wrote:
[snip]
> If you have the memory, why not make a RAM disk to store any
> local information (I guess it depends on the size of the database).
> However, have you seen the new motherboards from Tyan and
> Supermicro that can handle up to 16 Gigs of DDR!? Very cool
> possibilities now.

An interesting idea. One issue I have, however, is the need to BLAST
genomic vs genomic without masking for repetitive sequence. BLAST
tackles this by storing a huge number of alignments and often delivers
80MB+ result files. Sure, I can parse the result on the fly before
sending it back to the master node, but I have never paid attention to
RAM usage under these circumstances. I will make a note to check during
the next batch. In any event, RAM may be cheaper now, but a 40GB IDE or
even high end SCSI is far, far cheaper than 16GB of DDR.

--
Eric Engelhard - www.cvbig.org - www.sagresdiscovery.com