Folding At Home
Robert Bradley has proposed that we donate our idle CPU time to the Folding@home distributed computing project. Seems like just the thing to do for the new Babylon Cluster while in-between jobs. Indeed, we now have 25 CPUs on the cluster doing protein folding calculations, as well as two dual-CPU Power Mac G5s at the lab, and more lab machines to come. Perhaps Team Holmes will add a drop to the bucket, but certainly not as much as the 1337 overclocking crews dominating the top tier.
-- Andrew Uzilov - 03 Feb 2006
Folding@home is a great project, and perfect as use for idle cycles during setup, but once the cluster is running I am more inclined to fill up our compute queue with jobs of our own. There are plenty of CPU-intensive tasks that we can queue up to address questions in evolutionary genomics that better fit our research priorities. Considering the amount of comparative sequence data out there and the CPU-hungry nature of some of our algorithms (particularly MCMC ones, but also alignment, grammar training and parsing, genome annotation, rendering, etc.) we should have no problem filling idle cycles. (In fact, if Sun Grid Engine is now working OK, I have plenty of jobs ready to queue up more-or-less immediately.)
-- Ian Holmes - 06 Feb 2006
Not a problem. Folding@Home is just something to occupy the cluster time with until our own calculations are launched. Folding@Home processes are scheduled with the lowest priority possible, so they will only occupy as much CPU time as isn't actually being used. So, if we launch any SGE jobs, the Folding@Home processes will not steal CPU cycles away from our own calculations - they will just be paused until CPU time becomes available again. But if that's still unsatisfactory, we could kill the Folding@Home jobs completely, once our own calculations get started.
-- Andrew Uzilov - 06 Feb 2006
Cool and the proverbial gang. Let's aim for zero spare cycles ;-) -- Ian Holmes - 06 Feb 2006