Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!ncrlnk!ncrcae!hubcap!boulder!foobar!grunwald
From: boulder!foobar!grunwald@ncar.UCAR.EDU (Dirk Grunwald)
Newsgroups: comp.parallel
Subject: Re: Subdivision (was: scalability of n-cubes, meshes)
Message-ID: <7485@hubcap.clemson.edu>
Date: 19 Dec 89 14:22:39 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 14
Approved: parallel@hubcap.clemson.edu

>>>>> On 14 Dec 89 15:46:07 GMT, birenboi%girtab.usc.edu@usc.edu (Aaron Birenboim) said:

Aaron> Were people keeping the flit routing in mind when they made the
Aaron> claum that a return to hypercubes would be a good decision for
Aaron> optical interconnection networks?  If so, why?  The flit routing
Aaron> will take a similar amount of time, the messagge transmission
Aaron> is dependant on simple inter-node bandwidth, number of hops
Aaron> do not have such an extreme effect.
Aaron> --
-----

The reason for moving to tori/cubes is bandwidth; if you have more
bandwidth than you know what to do with (i.e. optics), you'd go back
to a lower number of hops (i.e hypercubes).