[Beowulf] torus versus (fat) tree topologies

[Chris Sideroff]

Greg Lindahl wrote:

> On Mon, Nov 08, 2004 at 11:45:12PM -0500, Chris Sideroff wrote:
> 
> 
>>  I have had the luxury of testing an SCI 2D torus cluster and found 
>>latency performance to be exceptional.
> 
> 
> Right. So is that a feature of the NIC, or a feature of the torus
> instead of a fat tree/CLOS/CBB network? Answer: the NIC.
> 
> There are several examples of low-latency fat tree/CLOS/CBB networks.

    I realize that, which is why I was looking for other comments about 
torus vs a (fat) tree networks.  I suppose that sentence implied that 
the low latency I seen was solely from the torus configuration - which 
was not the intention.  It was in reply to a preivous comment of the low 
latency of fat tree connects which is also obtainable in torus networks 
- essentially what you are saying :-)

   As a side note, since my original post about this the Nov 2004 Top500 
has come out.  I did not realize the dominance that myrinet has on the 
list - #1 with 38.6% share - while no other proprietary high-speed 
interconnect is even close.

Chris Sideroff