Load balancing in distributed communication systems: performance evaluation with SWN models

Modelling becomes more and more important in computer science. Recent systems which are in general composed of several resources of different nature involved in sharing tasks sent by users need high level models to be studied. In theses complex systems with several interactions, load balancing between the servers which process the requests is necessary if performance is required. Many works have been done on load balancing, however none of theses studies uses the stochastic well formed Petri nets (SWN) which offer a high level modelling. In this work, we are interested in the adaptation of an algorithm of load balancing with SWN in a distributed database. Our study takes into account the synchronization, the identity of the resources and their cooperation. So as to point out the impact of the load balancing on performance, we propose two models: a model with load balancing, another one without load balancing. Performance measures are carried out using symbolic simulation.