Efficient parallel simulation of a sliding window protocol

Abstract

Two parallel algorithms for simulating a sliding window communication protocol are described. The first exploits the parallel generation of interarrival, service and transfer times, whilst the second is based on parallel prefix computations. Both algorithms allow the processors to work largely asynchronously with each other. The efficiency of the two approaches is evaluated experimentally, using an implementation on a cluster of 12 processors connected by fast Ethernet. © 2002 Elsevier Science B.V. All rights reserved.