Performance Analysis Of Polling System With Multiple Gated Service

As an important control model, polling systems are used in extensive areas such as industrial control, communication systems, computer networks, production and traffic systems, etc. In recent decades, a lot of research and analysis based on polling model have been continuously developed to extend their application areas. As a model of practical use and research, polling systems have played an important part, especially in communication systems and computer networks, in which the types of the medium access control are mainly divided into polling multiple access and random multiple access. The random access systems are widely and flexibly used in distributed network structures because the random access systems do not need center control nodes. However, it is hard to maintain the network performance only by the means of random access and collision resolution when the systems have heavy loads. In this case, polling multiple access is more superior than the random multiple access.A model of polling systems usually consists of one central server and terminals, its processing structure consists of the arrival process of message in each terminal, switching process between the terminals and serving process of the message. Polling systems can be sorted into Limited ,Gated and Exhaustive discipline systems according to the kinds of the service disciplines. There are three major performance measures used in analysis of polling systems , including mean queue length, mean cyclic time and mean delay of the message, etc. For a long time, a lot of researchers have struggled to improve the performance of the polling systems by the studying polling order, service disciplines and servicing order in the terminals. The choice of service disciplines decides the servicing time and proficiency of each terminal, it is necessary to focus on both priority of service demand and fairness. Therefore, study of polling servicing disciplines is a focus in these years. NExact analysis of polling systems forms the basis of theory and application. By the means of Markov chain theory and multiple probability generating function, this paper sets up Limited ,Gated and Exhaustive models of polling systems in discrete time. Basing on these models, exact analysis of mean queue length, cyclic time, and message delay are given, then, comparisons based on these performance parameters are also given, which builds up the theoretical foundation of consecutive research about polling system.Study on service disciplines of polling system is continuously attached importance to because of its extensive application value. Based on the emphasis of and Limited, Gated and Exhaustive polling systems, a model of polling system with multiple gated discipline is proposed. By the means of multiple probability generating function and embedded Markov chain theory, we have constructed the model of the system, gave the derived process of mean queue length, cyclic time and message delay in detail, and obtained exact expressions of the performance parameter. The theoretical expressions correspond well with the computer simulations, then, some analysis and discussion are given and some helpful conclusions are given, which provides theoretical basis in terms of the construction of the practical polling system, moreover, serving disciplines of polling system are extended than the previous serving disciplines.This paper is organized as follows:In section 1, we introduce polling system in detail, and propose the existing problem what we are going to solve.In section 2, we introduces three basic polling systems with various service disciplines: Limited-K, Gated and Exhaustive service disciplines. By the means of embedded Markov chain theory and multiple probability generation function, analysis and comparison based on the performance of there various polling systems are given.In section 3, we propose a model of polling system with multiple gated service, based on the existing Gated, Exhaustive and Limited polling systems. By using embedded Markov chain theory and multiple probability generation function, we obtain the exact expressions of system performance.In section 4, computer simulations correspond well with the theoretical expressions. Based on these data, we have lots of analysis and obtain some helpful conclusions.In section 5, we makes a conclusion about this paper, account for the achievement and existing issues in the research. The prospect of the further research is mentioned in this section too.