Optimal resource allocation for a heavy BLASTN run?

[DNA Sorcerer]

Hello,

I need to run BLASTN (locally) on my institution's cluster. My query contains a million short reads and DB contains an equivalent number of sequences.

What would make the run go faster? More threads, more RAM or both? What would be a number of cores and ram beyond which no benefits are gained?

Thanks in advance,

[Ciaran]

RAM usage will be dependent on the database size, so extra RAM will not speed up the computation.

Multi-threading may help as long as the your blast call is consistent with your resource request.

I have my reservations however with how well blast implements multi-threading, i generally split my input files and run separate blast calls before catting the output files, this however maybe an artifact of my system set up.

Ciaran

[GenoMax]

You are going to have to do some experimentation with your cluster since each cluster is setup differently, is likely to have limits on how you use the resources and how much of them are going to be available to you (unless you are the only user).

See this recent thread about efficient ways of submitting blast jobs. The thread focuses on SGE but if your cluster users a different queue management system then you will need to adjust things accordingly. http://seqanswers.com/forums/showthread.php?t=29251