On Sep 16, 2008, at 12:07 PM, Ethan Mallove wrote:
What happens if one uses --local-scratch, but leaves out the --scratch option? In this case, I think MTT should assume --scratch is the same as --local-scratch.
In this case, my $0.02 is that it should be an error. --scratch implies a --local-scatch, but I don't think the implication should go the other way.
Could the "local" in --local-scratch ever be misleading? Couldn't a user ever use a mounted filesystem that's faster than all their local filesystem? Should it be --fast-scratch?
Mmm... good point. What if we name it what it really is: --mpi- install-scratch? This also opens the door for other phase scratches if we ever want them. And it keeps everything consistent and simple (from the user's point of view).
For future work, how about --scratch taking a (CSV?) list of scratch directories. MTT then does a quick check for which is the fastest filesystem (if such a check is possible/feasible), and proceeds accordingly. That is, doing everything it possible can in a fast scratch (builds, writing out metadata, etc.), and installing the MPI(s) into the slow mounted scratch. Would this be possible?
Hmm. I'm not quite sure how that would work -- "fastest" is a hard thing to determine. What is "fastest" at this moment may be "slowest" 2 minutes from now, for example.
I'm looking at the patch in detail now... sorry for the delay... -- Jeff Squyres Cisco Systems
