[Beowulf] Glenn Lockwood's Thoughts on the NSF Future Directions Interim Report

[Jason Riedy]

And Michael Di Domenico writes:
> seems to me [NSF] needs to get back into building the HPC community of
> PEOPLE rather then building hero machines at six or seven
> installations across the us.

Note that many recent calls include community-building aspects, so I
believe that is recognized.

The NSF also is looking for applications that are not well-served by
the current vendor solutions.  IMHO, those solutions are trending
towards being opaque boxes or "local clouds", where power &
environmental are managed locally but the system management could
well be outsourced (e.g. a typical cluster, a BlueGene, a Cray XC,
etc.).  If the major computing centers figure out data aspects, then
there's one fewer reason for those black boxes even to be
distributed.  When system management is mostly automated, and we all
know it could be if we stop standing on each others' toes, people
managing the hardware really don't learn anything.  Think Google in
Lenoir.

But for *new* platforms...  That's where you need to distribute the
expected chance of failure across many smaller locations and need to
build off their expertise...