On Tue, 15 Apr 2003, Jerome BENOIT wrote:
> How may we configure LAM on a Mosix cluster ?
Somehow I missed this post -- sorry! :-)
LAM doesn't play too well with MOSIX. Since LAM opens dedicated
communication channels to processes, if a process suddenly moves, then
communication channels may break. :-(
There's probably a way to make LAM be able to run nicely on MOSIX clusters
(especially since we now have basic checkpoint/restart capabilities -- I'm
guessing that MOSIX can somehow tell processes "hey, you're about to
move", which would reduce the problem to be the same as
checkpoint/migrate), but it's not something that we are currently looking
into.
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
