Basic HTML version of Foils prepared 21 October 1995

Foil 62 Hybrid Distributed/Shared Memory Architectures

From New CPS615 Foils-- B 28 August 95 CPS615 Basic Simulation Track for Computational Science -- Fall Semester 95. by Geoffrey C. Fox


1 Consider "nodes" = printed circuit boards (Paragon) or bunch of printed circuit boards sharing a BUS (SGI Challenge) where the N nodes are connected as a distributed memory architecture
2 Inside each node, may have n=2,4,8 .... processors sharing a single memory.
3 Should one use this shared memory or just consider total machine as distributed with Nn nodes?
4 This is seen today (1995) as a network of SGI Challenge systems (i.e. metacomputer where conventional HIPPI network joins set of shared memory machines) or as a specialized torus network connecting Paragon nodes consisting of multiple i860's

in Table To:


© on Tue Oct 7 1997