[Bioclusters] LSF vs SGE

[Steve Pittard]

LSF vs. SGE

We've been using Platform LSF to manage our BLAST/HMMSEARCH
cluster for the past year. We've made the cut over
to Sun Grid Engine and are currently completing stress testing 
after which we will use SGE exclusively. Until then we 
are running LSF and SGE in parallel (no pun intended) though 
our underlying Perl programs refer to enviroment varibles
making it possible for the cluster to use one or the
packages with just a quick change. In this fashion we
can get back quickly to LSF if SGE is having problems.

I think LSF is an excellent product with lots of tunability (
is that a word ?) and I'm pleased with their support of the
product. None of my questions to Platform tech support have
gone unanswered. On the other hand I wasn't at all pleased 
last year when we shelled out a premium price per CPU only
to find that a mere two months later the price had dropped
to a quarter of what we paid. Frankly I felt price gouged.
Bottom line - make sure you are getting the best price
possible if you go with LSF. If you are buying from 
a cluster provider (e.g. MiroWay) they usually negotiate
the best price for your environment but make sure you check 
it out.

SGE is doing the job thus far and most of the questions I've
had were answerable via the internet and the support forum.
If money weren't an issue (when isn't it ?) I would renew
LSF since we've gotten good performance data from it and
know what to expect in terms of functionality. But since 
budget money is tight this year I'm compelled to use free 
software where possible.


Steve Pittard	 | http://catalina.bimcore.emory.edu (HOME PAGE)
Emory University | wsp@emory.edu, wsp@bimcore.emory.edu  (INTERNET) 
BIMCORE Support	 | 404 727 0038