Re: Approximating Hypothetical Beowulfs

Brian Atkins (brian@posthuman.com)
Wed, 02 Jun 1999 21:14:46 -0400

Chandra Patel wrote:
>
> Hello list!
>
> I'm trying to develop a formula to predict the Linpack benchmark rating of a
> Beowulf. To make it easy I assume all nodes have the same processor type and
> the same bus and memory specifications. The factors I've identified so far
> are:
>
> 1) processor type (speed, flops rating, etc.)
> 2) memory (subfactors include memory access time, cache speed and size)
> 3) Ethernet capacity (throughput)

Yeah besides capacity and architecture (as someone else pointed out), I think you might also want to take into account the latency of the different network technologies. I think some Beowulfs use special types of networking (non-ethernet?) in order to provide faster message passing.

-- 
Join the ExI/>H SETI team today:
http://setiathome.ssl.berkeley.edu/cgi-bin/cgi?cmd=team_show&id=393
-------------------------------------------------------------------
"It IS possible." - Vincent, _Gattaca_