EFFICIENCY ESTIMATION OF COMPUTING GRIDS WITH VARIOUS TRAFFIC TYPES

Abstract

For the efficiency evaluation of computing grids with different types of traffic, reenterable colored Petri nets are suggested. The principles of constructing models in the form of reenterable colored Petri nets are presented. The advantages of the proposed type of constructing models are shown on an example of a rectangular computing grid with an arbitrary size. The number of vertices for different types of models is estimated. The behavior of the real-size computing grid model is studied for various traffic intensities and distribution functions: Poisson, normal, and Rayleigh. The results of estimating the average packet delivery time and the performance are given under the workload conditions. The adequacy of the constructed reenterable models of computing grids to the models of direct mapping is confirmed. The efficiency of computing grids with conditions of workload and malicious traffic is investigated; the results of computing experiments with different kinds of malicious traffic are confirmed.