Re: Dual Ethernet Slowness Causing Lousy Benchmarks? (was: A very

Walter Lundby (lundbys@xnet.com)
Tue, 06 Jul 1999 23:24:23 -0500


Dan Hollis wrote:
>
[...]
> I would imagine tcp scaling is more important to the clustering people.
>
> -Dan
>

I have some questions here for the community. I am considering a complex
sniffer with token ring, and at least two ethernet nics. A third nic
would provide access to the company lan {inquiry and control}. We need
to measure timing between messages traversing various platforms and try
to relate them for verification of performance models. I would prefer
one platform with several nics as the timebase would be common. Accuracy
in the milli-second range is fine. The machine would do nothing but
collect the messages.

The questions:

1) Will this performance problem impact timing measurement or cause loss
of data when more than two nics are being measured.

2) Would a multi-processor motherboard help?

3) Should I vary the model/brand of nics used so that I avoid problems
relating to common registers and interrupts? {Suggestions}

4) Would Real-Time linux be a better solution or just buy more problems?

Comments are appreciated.

Walter Lundby
lundbys@xnet.com
or
cwl001@email.mot.com

-
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/