[Beowulf] Article on Sanger centre
hahn at physics.mcmaster.ca
Tue Jun 7 08:04:37 PDT 2005
> Slashdot is carrying an article on the infrastructure at the Sanger
> Goodness - still running many Alpha processors.
hey! what's wrong with alphas?
I was rather surprised that they've left it until now to consider replacing
the alphas, though. I wouldn't have thought that blast-like stuff is all
that great on alphas. perhaps they do more MD-type stuff. it's also
gratifying to see a real bioinformatics site talking fairly modest numbers
when it comes to sequence DB's. (so often, it seems like bioinformatics folk
get carried away into extrapolating into the many-PB range...)
> Might be worth reversing the car up to the loading bay the day they
> are got rid of....
we have a much smaller number (around 450), and are planning to redeploy
then, rather than getting rid of them. we've had them running (in 100-150p
clusters) for 4+ years, and are buying >6K mostly opterons. but the alphas
will be coalesced into a single cluster that will still be interesting for
our es40's dissipate about 750W, and I'd say that a single modern
dual-opteron (say 240W) is modestly faster in compute power. there's
definitely a power savings there, but also a drastic space savings.
but there are situations where power and space is relatively cheap...
More information about the Beowulf