FNN vs GigabitEther & Myrinet

Eray Ozkural (exa) erayo at cs.bilkent.edu.tr
Wed Oct 24 04:29:57 PDT 2001

Hash: SHA1

On Wednesday 24 October 2001 05:50, you wrote:

> I have never seen a GigE NIC pushing ~1 Gb/s. It's hard to push so much
> with IP, as the IP stack becomes quickly the real bottleneck. To reach the
> full hardware performance, you need an OS-bypass protocol, or a zero-copy
> IP stack. Folks from the project project at Duke used a such IP stack in
> FreeBSD to reach 2 Gb/s TCP over Myrinet. Without this special kernel and
> driver, the same hardware provides 1.2 Gb/s TCP with the Linux 2.4 IP
> stack. It was way less efficient with 2.2 kernels.

Which brings to my mind GAMMA project. Why wasn't that project backed up by 
other people so that it could be made into production use? The latencies 
reported in the papers and the website are absolutely awesome. With GAMMA/MPI 
we could easily have the courage to play with fine-grained algorithms.

That would surely beat LAM/MPI over TCP/IP.


- -- 
Eray Ozkural (exa) <erayo at cs.bilkent.edu.tr>
Comp. Sci. Dept., Bilkent University, Ankara
www: http://www.cs.bilkent.edu.tr/~erayo
GPG public key fingerprint: 360C 852F 88B0 A745 F31B  EA0F 7C07 AE16 874D 539C
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org


More information about the Beowulf mailing list