[Beowulf] using two separate networks for different data streams
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.
Keith D. Underwood kdunder at sandia.govMon Jan 30 06:32:37 PST 2006
- Previous message: [Beowulf] using two separate networks for different data streams
- Next message: [Beowulf] using two separate networks for different data streams
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> The Cray XT3 is an example, I think, of a system which sends > non-IB-like packets over an IB physical layer. But I don't think the > details are public and I don't know much about the XT3. The only heritage that the XT3 shares with IB is that IB was a strong (but by no means exclusive) driver of the development of high-speed- signaling technology. The physical layer on a link by link basis doesn't even have the same signaling rate as IB (3.2 Gb/s/wire pair for XT3 vs. 2.5 Gb/s/wire pair for IB). XT3 is 12 bits wide and I'm not sure if anybody has done a 12X IB yet. Most importantly, XT3 has a physical layer reliability protocol with retransmit on the links between the routers so that end-to-end messaging is extremely reliable. The requirement was for them to have no more than a 10^-21 BER host-to-host and so the physical links between routers are much better than that. Keith -- Keith D. Underwood Scalable Computing Systems Senior Member of Technical Staff Sandia National Laboratories kdunder at sandia.gov
- Previous message: [Beowulf] using two separate networks for different data streams
- Next message: [Beowulf] using two separate networks for different data streams
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
