Re: Approximating Hypothetical Beowulfs

From: Dan Clemmensen (Dan@Clemmensen.ShireNet.com)
Date: Fri Jun 04 1999 - 18:14:40 MDT


Eugene Leitl wrote:
> You can achieve about ~20 us messaging latency for small messages
> if you use M-VIA on (DEC Tulip) FastEthernet. Vanilla Linux TCP/IP gives
> you ~60 us. A cut-through switch introduces a few us overhead. With
> the channel bonding kernel patch one can coalesce up to 4 NICs into
> one channel, all of them running full-duplex.
>
I'm fairly sure (gut feel, not mathematical analysis) that you
can make much more cost-effective use of a 4-NIC solution by
establishing a four-dimensional hypergrid than by bonding.

Of course, it depends on the problem.



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:03:59 MST