lkml.org 
[lkml]   [1998]   [Nov]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: SMP scalability: 8 -> 32 CPUs
From
Date
>>>>> "Larry" == Larry McVoy <lm@bitmover.com> writes:

[snip]

Larry> What does this all mean? It means that if you try and scale
Larry> beyond a handful of processors, you complicate your kernel
Larry> substantially and make it almost impossible to keep it a
Larry> lightweight thing on a UP (look at Sun vs Linux lmbench numbers
Larry> on the same hardware, that's all the proof you need).

Larry> Furthermore, most people have UP or 2 way SMP machines, very
Larry> very few people have 8 way SMPs and even fewer have 32 way
Larry> SMPs. And that distribution will remain quite constant.

Larry> So ask yourself - should the Linux kernel team do an incredible
Larry> amount of work, destroying performance and reliability in the
Larry> overwhelmingly common case, for the benefit of a very, very
Larry> small number of machines?

I cut out a fair bit of your posting ... though I would like to
comment on the general stuff.

I agree with you that the market is primarily in the 1-2 CPU boxes at
the moment, but I believe we'll see Intel and others trying to push
2-4 way SMP machines soon for end users soon (I just don't believe
they can continue to boost the performance curve). Besides, several
companies have announced/delivers 8 way x86 SMP boxes now.

Linux is beginning to move into the low/mid-range server market as PCs
are beginning to become interesting when looking at I/O performance,
at least that is the trend I see here at CERN. Until now we have
primarily been using PC's for number crunching, but with the latest
chipsets from Intel it has started to become interesting to use them
as I/O servers (tape/disk) as well.

Right now our filesystem and SCSI (and probably network stack as well)
code just doesn't scale for this but I really want to see that change
in 2.3.x (I hope and believe I am not the only one with that vision
;-). I would sure like to be able to replace Sun 450 & DEC 1200 class
machines[1] with Linux boxes, but right now it is just not feasible.

So basically what I wanted to point out is that we need more
scalability to take over some of the new markets where Linux is
beginning to evolve. I think these are very important areas and it is
important not just to focus on private desktop users (of course we
should not forget this market either).

Jes

[1]: This is my own personal dream, not an official company statement!

-
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/

\
 
 \ /
  Last update: 2005-03-22 13:45    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site