Share Email Print

Proceedings Paper

What do you do when the network is 1 to 2 orders of magnitude faster than the backplane bus?
Author(s): Seth J. Abrahams
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

The prototype high performance switching system will connect several different types of machines with widely varying clock and internal bus speeds. At one end of the range we have a CRAY YMP and a CRAY XMP with a high performance parallel interface (HIPPI). At the other end there are many SUN 3/260 VME bus machines. In between there are various flavors of MAC IIs. This presents a substantial speed matching problem. The CRAY HIPPI channel speed is 100 MB/s. The SUN VME bus speed is approximately 1 MB/s with bursts to 13 MB/s, the MAC II NU-BUS obtains about the same speed. So what do you do when the network is 2 to 3 orders of magnitude faster than the internal bus? The short answer is use big packets and even bigger buffers. This paper expands on that theme and describes the approach LLNL and Ancor Communications took in the case of the SUN VMEbus machine.

Paper Details

Date Published: 1 March 1992
PDF: 9 pages
Proc. SPIE 1577, High-Speed Fiber Networks and Channels, (1 March 1992); doi: 10.1117/12.134933
Show Author Affiliations
Seth J. Abrahams, Lawrence Livermore National Lab. (United States)

Published in SPIE Proceedings Vol. 1577:
High-Speed Fiber Networks and Channels
Kadiresan Annamalai, Editor(s)

© SPIE. Terms of Use
Back to Top