[ale] mosix clusters?

[Jeff Hubbs]

On Mon, 2002-06-24 at 01:25, Joseph A Knapka wrote: 
> Jeff Hubbs wrote:
> > 
> > That's what I would assume as well.  I don't have the 'fu to know just
> > how much context-switching overhead that would entail.  I do wonder,
> > though, if there might be some "sweet spot" that takes advantage of the
> > way the processes use the CPU/mobo architecture, what with instruction
> > pipelining and L1/L2 cache.
> 
> That is a very interesting question. After thinking
> about it a bit, I believe it's possible to have
> multiple tasks running in such a way that CPU
> cache invalidations and refreshes happen in a
> particularly "good" sequence that will actually
> allow each of several identical compute-bound processes 
> to make progress more quickly when run simultaneously
> than when run serially. However, I don't think (with
> current kernel technology) that you can plan
> for this, and in general it wouldn't happen. You'd
> have to get lucky. If you got unlucky, which you'd
> normally do, then everything would run slower
> than in the serial case.
> 
> -- Joe
Joe - 

I think that you're right about not being able to plan for this.  What
you'd have to rely on is the asynchronous nature of the various
processes.  It'd be like two people shooting shotguns at two targets
such that the shot paths cross at the same point in the air; some of the
shot from each gun would still make it to the target.  


---
This message has been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.