A SEMI-MARKOV MODEL FOR THE PERFORMANCE EVALUATION OF MULTIPROCESSORS

Abstract

Due to the complexity involved in the design and implementation of multiprocessor systems, the trend to develop the analytical models with some simplifying assumptions is gaining popularity over the last decade. A simple yet efficient technique to model a multiprocessor system is to use semi-Markov process (SMP), to describe the behaviour of each processing element (PE), independently. The system with N PEs can then be modelled using N such SMPs, with the assumption that all PEs behave identically. Solving the model requires only one SMP to be considered. This technique:! dramatically reduces the solution complexity. In this dissertation, a discrete time model for the memory interf