Hi,
> > [... stuff deleted ...]
> > A more reasonable estimate hovers closer to 4 when discussing
> > 2.2.x. In my opinion, one or more of these hot-to-trot
> > big iron houses (IBM, SGI, Sun) needs to put more actions to
> > their words and contribute some useful code in the way of
> > scalability. Chances are good none of them will, as
> > highly-scalable systems running commercial Un*x implementations
> > are their honeypot du jour.
Th point is, that in almost every case, "scalability" is not achieved by
"adding some code/function" to the kernel. Most of the time, it's about
carefully using locks in important kernel structures to allow kernel
functions accessing them concurrently. In Linux, these datastructures
are implemented quite differently compared to other unices. Although
most unices behave similar from a userland perspective, the grunt
stuff in the kernel is often quite different. Doing things in the kernel
that scale well on lets say above 8 CPU's is hard. Providing
the very same functionality scaling on 16, 32 or even 64 CPU's is
a whole *different* story. I wish - for example - Sun would release
the SunOS 5 ( aka Solaris ) kernel sources they use on an E10000
( 64 CPU UltraSPARC - big iron in it's very sense ). By just *looking*
at the code, it would reveal *dramatic* differences to what used to
be *standard Unix kernel algorithms* a la Maurice Bach ( "The design
of the Unix operating system" - a must-read ).
Although I am convinced,. Linux will evolve to a point were it will
scale on more CPU's than today, I am also convinced, it will look
quite differently in the internals.
Please keep in mind, that it took a long time for *all* of the big-iron
companies to get their Unix scaling on large SMP machines. They
used big engineering teams of very competent programmers to get
what they have today.
Linux *will* get to this point - but it will take time. Personally, I am
very happy about all the valuable pieces of code coming from
various sources moving Linux in that direction.
<humor>
To quote Jay Estabrook, a couple of postings ago:
"YOU could make a difference! Don't wait for other folks to do it..... "
8-)
</humor>
Have fun !
Kirk
-- ------------------------------------------------------------------------------- Thomas Weyergraf kirk@colinet.de-- To unsubscribe: send e-mail to axp-list-request@redhat.com with 'unsubscribe' as the subject. Do not send it to axp-list@redhat.com
This archive was generated by hypermail version 2a22 on Tue Feb 1 05:33:31 2000 PST
Send any problems or questions about this archive to webmaster@alphalinux.org.