Poll: distributed sequence computing

[bchaffin]

Hi folks,

As has been mentioned in a couple other threads, I've been working on a system to pursue all unreserved sequences <1M in parallel across many cores and machines. I've also been thinking about whether and how to make this more widely available, and I wanted to get your opinions on the best way to proceed.

One relatively simple option is just to publish my scripts. Assuming people are interested, we'd divide up the <1M space using a reservations thread, and each contributor would work all the unreserved sequences in their range. Today my Perl scripts are Linux-only and require that all machines share a filesystem. (A Windows port is probably not too hard.) They generally require some hand-holding if things go wrong, like network outages, disk full, etc., but are not really any harder to set up or use than any of our other tools.

Another option is BOINC. I think a BOINC server is feasible, though obviously it's a significant undertaking to establish and host it. Besides the work, one downside is that people who borrow unused cycles on machines they don't administrate (like me) may not be able to install the BOINC client. On the other hand, in theory people outside this forum could also easily participate (but who knows if anyone would). And of course it would avoid the overhead of a reservations thread and be a lot more tolerant of occasional participation.

Option zero, of course, is that we do neither of these and continue as is, which obviously works pretty well already.

What do you folks think? Would you participate in such a project, either running scripts or through BOINC? Which option do you prefer?

Thanks in advance for your advice.

ben

[schickel]

I copied this to a poll above, please vote there....