> > I'd like to ask (and then discuss) some questions:
> >
> > 1)Is distributed IPC a useful abstraction?
> > 2)Is distributed shared memory a necessary part of that abstraction?
> > 3)Does this of necessity _have_ to be part of the kernel?
>
> Yes, yes, yes.
Interprocess communication seems a useful abstraction. But I fail to
see by which leap of logic it follows that DSM is then a *necessary*
(in the mathematical sense of "necessary") part of that abstraction.
It might hypothetically be desirable, for implementation reasons, to
consider it a part of that abstraction. But that would be choice, not
necessity.
> Let's say you have 3 choices:
>
> 1. 95% use of a single machine
> 2. 80% use of a 16-way cluster, but 2 weeks development time
> 3. 60% use of a 16-way cluster, but 2 hours development time
Is there hard data to back these numbers up, or are they pulled from
the hat? Is it a desire or a fact that DSM is a simple solution?
> Since you didn't answer this, I'll do it for you: yes. People often
> need to reuse code and skills. Human time is limited,
However, if (for the sake of argument) DSM is the wrong abstraction,
coding to it and then needing to rework the design to make use of DSM
while at the same time trying to factor out the assumption that
memory is really shared and distributed would be a waste of human
time.
--
,_
/_) /| /
/ i e t e r / |/ a g e l
-
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/