CONGESTION CONTROL IN ATM NETWORKS- A NEW FUZZY SYSTEM BASED QUEUE MANAGEMENT SCHEME

Abstract

In Asynchronous Transfer ' Mode(ATM) networks, to increase the bandwidth_ utilization, statistical multiplexing can be used effectively. However, while achieving this result, one has to accept congestion as an unavoidable nuisance. To minimize congestion, a new queue management scheme based on fuzzy logic, is proposed in this thesis, wherein buffer occupancy as well as burst length are represented by -fuzzy sets.;n the new scheme buffer occupancy as well as burst length are represented by triangular membership functions.From these membership functions, the degree of membership, to a particular set, associated with 'buffer occupancy and burst length, can- be read. This process is called fuzzification. From these membership values and corresponding sets, blocking, to the incoming burst of cells, to be offered, in fuzzy terms, can be found out. This process is called rule based- inference. Then by applying suitable defuzzification method such as min-max or max-min defuzzification the percentage blocking to be offered, to the incoming burst of cells, at that particular buffer occupancy and burst length can be determined. ' I The performance . of the new scheme has been compared with that of the fixed threshold scheme. It has been shown that the new scheme improves the performance with reference to major. parameters namely - -throughput, average delay and dropped cells(ce11 loss probability). The improvement in performance of the suggested scheme is achieved at the cost of higher time complexity as shown in the thesis. This scheme is also far more flexible than the fixed threshold scheme.