Thursday, April 15, 2010

MPI, and parallel jobs, and conference roundup

The morning session on parrallel processing was mostly updates on the progress of getting MPI working reliably. John Walsh was talking about the task force, and how the number of sites that publish that htey support MPI has gone down slightly - but the number of sites where it actually works has gone up. The monitoring processes checking MPI jobs has helped significantly with that.

The next talk was about where to next, and giving the users the ability to control more things - there are several different categories of parallel jobs, and it was good to see explicit consideration being given to them. There's not a fixed timescale mentioned for implementation of these - during the immediate transition, there'll be a focus on documentation (rather than changing things); which is a very sensible plan.

Parallel tasks are the big area for us, where we get the most request that we can't quite do; I think we'll be working in this area for a while.

That's the last session of the parallel sessions [0], so everything it starting to close down now, before the closing plenary.

Overall it's been a good conference; with lots of cross talk between users and techy types, which is the the key thing that doesn't happen elsewhere. Even with my flights home being cancelled (looks like it's going to be a long weekend ... and not in the holiday sense!), it's been good to be here.

I think the key theme running through the conference is to do with the end of EGEE, and the changes. With then general sentiment being, "All that good stuff? That's not going to change", I hope the transition is as smooth as it looks like it will be.

[0] Pun intended, of course.

No comments: