Performance of Shared Memory in a Parallel Computer

  • Authors:

  • K. Donovan

  • Affiliations:

  • -

  • Venue:
  • IEEE Transactions on Parallel and Distributed Systems
  • Year:
  • 1991

Quantified Score

Hi-index 0.00

Visualization

Abstract

A method for analyzing the lengths of memory queues when the network is conflict-free isdescribed. An algorithm based on this method is shown to efficiently determine the upperand lower bounds of the queue length. Analysis indicates that the strategy of usinghashing to spread data across memory modules is a good one. Results show that if thesize of the system is increased while maintaining a constant ratio of numbers ofprocessors to memories, then, asymptotically, the slowdown in performance from conflictsat the memory modules is Theta (log m/log log m). For m and n less than 100000 andlambda between 0.25 and 4.0, the graphical data confirm this growth rate.