Distributed Performance Improvement of Alternating Iterative Methods Running on Master-Worker Paradigm with MPI

Abstract: In this paper we present methods that can improve numerical parallel implementation using a scalability-related measure called input file affinity measure on 2-D Telegraph Equation running on a master-worker paradigm. The implementation was carried out using Message Passing Interface (MPI). The Telegraph Equation was discretized using Alternating Direction Implicit (ADI), Iterative Alternating Direction Explicit Method by D'Yakonov (IADE-DY) and Mitchell-Fairweather Double Sweep Method (MF-DS). Parallel performance compare of these methods using the Input File Affinity Measure (Iaff) were experimentally evaluated and numerical results show their effectiveness and parallel performance. We show in this paper that, the input file affinity measure has scalability performance with less information to schedule tasks