English abstract
The new emerging technologies such as distributed web applications, video and audio on demand, video conferencing and internet technology are putting a tremendous strain on the networking infrastructure. Recent research efforts have addressed performance analysis of ATM networks from the pure low-level communication aspects without relating it to real applications. Distributed applications running over high-speed network architecture, introduce a number of performance factors that are not considered in the low-level communication models. Examples of such factors include database query distribution, query sizes, serving the queries, replies delays. The estimation of the response time is an important parameter in the performance study of a system. Another example is estimating the saturation level in the system. That is estimating the maximum request generation rate or the maximum number of clients in the system. Also it is helpful in studying the effectiveness of changing some parameters in the system. This study proposes an analytical model that captures the effect of ATM network on distributed client-server applications in terms of response time. The response time has several components such as propagation delay, processing delay and queuing delay. The proposed model of an ATM LAN identifies the queuing delay that exists in a distributed applications. First the system is divided into sub-systems and we model each sub-system individually then an aggregation of all the sub-systems lead to an overall model for the system. The idea of division in sub-models is very useful for bottlenecks' detection in the system. Furthermore, an appropriate Colored Petri Net models are developed and used for simulation and validation of the analytical results. The results obtained by the simulation were in close agreements with the analytical results
