Flat Network Neighbourhood (again) (was Re: 32-port gigabit switch)
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Ken Chase math at velocet.caThu Mar 6 19:22:46 PST 2003
- Previous message: 32-port gigabit switch
- Next message: 32-port gigabit switch
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Thu, Mar 06, 2003 at 11:40:45PM +0200, Eray Ozkural's all...
> Hi Mark!
>
> On Thursday 06 March 2003 21:55, Mark Hahn wrote:
> >
> > take 64 nodes, think of them as an 8x8 grid.
> > take 16 switches, think of them as 8 "row" switches and 8 "col" switches.
>
> Excuse me, is that 16 16-port switches? Wouldn't that cost a little too much?
>
> I've been thinking about novel static and dynamic networks for Beowulf
> systems, so I have a genuine interest in new architectures. A fat "dynamic"
> tree sounds like a nice idea.
>
> I'm not sure if your proposition is sensible, though. Surely, I agree that the
> "really cool switch" solutions are becoming too expensive....
are we just talking about FNN again? Why row and column? Isnt FNN
more efficient?
http://aggregate.org/FNN/
So for all these clusters, all nodes need to contact other nodes during
calculations, correct, which is why a full switch is needed?
/kc
--
Ken Chase, math at velocet.ca * Velocet Communications Inc. * Toronto, CANADA
- Previous message: 32-port gigabit switch
- Next message: 32-port gigabit switch
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
