Performance analysis of network access nodes using polling systems

Abstract

In recent years demand for high speed internet access and multimedia service has increased enormously. Therefore, to meet all these requirements the field of communication networks is evolving very fast. When we build up the communication networks infrastructure, first we must have to analyze the performance of each network component before deployment, to satisfy our valuable users. Queueing theory is one of the tools which can be used to analyze the performance of communication network components. In this thesis we analyzed the performance of the network access nodes to provide the desired quality of service. We used the queueing theory and proposed polling systems with different service policies to investigate the performance of network access nodes.<br />Markov chain is an efficient and powerful tool, which can be used to investigate the queueing systems. We exploited Markov chain techniques to develop our polling models. First we presented all the models with two queue systems and then generalized them to n-queue systems. There are multiple service classes which have different quality of service demands. Therefore we proposed models dealing with priorities as well.<br />First we performed the performance analysis of a polling system handling with preemptive priority service. In preemptive priority service model starvation is faced by low priority classes. Therefore we created a polling model for non-preemptive priority service discipline to tackle the monopolization of high priority services. We noted also the effect on performance by varying the value of switchover rate. After that we did the comparative study of both priority service traffic streams.<br />Later on, we presented polling systems with exhaustive service, round robin service with threshold triggered switchover and G-limited service discipline. We used the successive over relaxation method to solve all these systems. All the obtained performance characteristics of the mentioned polling systems were analyzed. Impact of switchover rate on the behavior of systems was also explored. In addition we did the comparison of these models. At the end we introduced a new decomposition approach to solve a nonexhaustive polling system with finite queues and studied its performance under varying situations.