Re: MPI vs PVM

[Junichi Uekawa <dancer@netfort.gr.jp>]

>     heiko> Some colleagues who have been at one of the last Super-Computer-Conferences
>     heiko> in Heidelberg told me that nobody was talking about PVM. PVM is almost
>     heiko> dead.
> 
> The other way to look at it is, since it works pretty well, there
> isn't anything to fix (extend, perhaps, but that is HARNESS, etc).
> 
> Does MPI perform well on a heterogenenous cluster :-)?

There is nothing in MPI that stops applications from running on 
a heterogeneous cluster, but there is nothing much that aids one.


My impression is that PVM is a old project that is dying a death.

We have two competing implementations of MPI (mpich and lam)
and they are both still alive, and scalability issues are
being worked on.

At supercomputing conferences, my impression is that PVM is no
longer actively developed, and MPI is trying to replace it.


Well, I only use MPI, so my views may be biased ;)



regards,
	junichi