[BiO BB] TimeLogic

[Chris Dwan (CCGB)]

>     Does anyone have any experience with TimeLogic's Decypher nodes?
> What kind of performance do the get and how do they compare in a real
> deployment to a cpu farm?

We (The University of Minnesota Center for Computational Genomics and
Bioinformatics) have two 'G' series DeCypher boards from TimeLogic.
We also have a total of 15 nodes of compute farm (various shapes and
sizes of Linux) scheduled using PBSPro and a bunch of home-brew
scripts. 

The TimeLogic machine is wicked fast, but it can be a little cranky to
work with.  There are lots of gotchas on query size (below 100,000bp
is best) and other things in the interface that just require learning.

Given that the user is willing to take the time to learn how to
drive it, people like it a lot.  For the folks  who have 80,000 to
100,000bp of sequence at a time that they need to TblastX vs. NCBI's
NT, dbEST, whole chromosomes, or whatever, it's a great machine.  For
those folks who want optimal results (Smith & Waterman and others...)
the TimeLogic machine can't be beat. 

The cluster, on the other hand, is most useful for 1st pass
annotation and analysis on massive sets of data:  millions of BLASTNs
to find the really close relations farm out very cleanly.  The cluster
also has superior  automation tools in place, but that's mostly
because we've been working on it longer.

I have no experience with ParaCell's products.

-Chris Dwan
 University of Minnesota
 Center for Computational Genomics and Bioinformatics