On Thu, 7 Sep 2000, Marty Fouts wrote:
> FWIW, large system scalability, especially NUMA is not tractable
> with a 'one size (algorithm) fits all' approach, and can be a
> significant test of the degree of modularity in your system.
> Different relative costs of access to the different levels of
> the memory hierarchy and different models of cache concurrency,
> especially, tend to make what works for system A be maximally
> pessimal for system B.
The first thing to tackle is making sure a lot of global
things in the system become per-node.
Something like that will increase scalability on various
kinds of hardware while not affecting performance adversely
on other systems.
Other scalability things will have to receive more care
as to not upset existing systems, but I'm sure we'll be
able to get a long way by simply moving things into local
structures...
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Sep 15 2000 - 21:00:10 EST