Re: OFFTOPIC: Regarding NT vs Linux
Darren Reed (darrenr@cyber.com.au)
Mon, 22 Sep 1997 08:50:49 +1000 (EST)
In some mail I received from Russell Coker - mailing lists account, sie wrote
>
> > Does anyone have any reports of a Linux system with 2 CPUs running
> > faster than a system with a single CPU?
>
> >A friend of mine has a 2xPii-266 and I was playing with it (booted of a
> >Jaz drive - its an NT machine) for a couple of hours.
>
> >Parallel kernel compiles are considerably faster. Network and disk
> >performance didn't seem to be noticably slower, but it is somewhat
> >difficult to tell without propery testing using a real setup.
>
> That's what I'm interested in, web and network performance (I'll assume
> that disk performance is only really limited by disk hardware). Having a
> "make -j2" run faster with 2 CPUs is no big deal. Having an NFS server run
> faster would be...
I'd be wary of relying on dual CPU systems with SMP for improved networking.
As a simple test I ran top in an xterm window (displayed on another box, over
the network) with a delay of 0. There were quite a few messages about
interrupts being missed, etc, although these appear to be network card
driver specific.
But, I think there are obvious answers to the question you're asking.
For example, if I'm running crack (just one instance) on a dual CPU
box that does NFS serving, you would at least expect there to be little
or no difference in NFS speed.
Until Linux has kernel threads, I can't see it being hugely faster for
NFS with dual CPU vs one CPU in server configurations where its CPU isn't
used a lot.