> On Sat, Mar 04, 2000 at 04:26:04AM -0500, Albert D. Cahalan wrote:
> > DSM isn't nonsense. Porting _is_ easier and computers usually _are_
> > cheaper than people. You pulled the "200" and "20" out of thin air;
> > it is actually "249" and "243".
>
> You want to buy a $20M computer and then avoid $1M in porting costs and
> you don't care about performance. Odd.
> Buy a $1M computer, make your code use MPI for $2M, and spend $5M on
> fast switches. Lower cost, higher performance. Or buy a real SMP machine
> for $20M.
>
> Or maintain DSM outside the base Linux kernel. And God bless you.
Unfortunately - it is closer to "buy a $20M computer and then avoid $40M
conversion cost". The personnel to do a large conversion is not always
available, or the time (1+ years) may not be available. Large applications
that use shared memory cannot be quickly converted. Usually it requires the
original author - and/or a fair sized team to make a detailed redesign to
do it, sometimes both author and team are needed. The time lost is also not
insignificant.
There are places for DSM. There are also places for CCNUMA implementations
of DSM. Just because it depends on underlying hardware is no reason to
"just say no" to something that is reasonable.
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil
Any opinions expressed are solely my own.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 21:00:27 EST