Author Description

In this thesis, we propose a novel self-adaptive fuzzy connection admission control (FCAC) scheme for multi-service packet switched networks. Currently, the full statistical characterization for online traffic is very difficult to get. However, many connection admission control (CAC) schemes are based on certain traffic models which limit their implementation into practice. This thesis proposes a measurement based CAC scheme without any assumption of source characterization. It defines a source characteristic factor which is derived from the peak rate, mean rate and QoS constraint to estimate the traffic behavior in terms of the possible loss ratio and bandwidth utilization. The estimation and measurement from actual traffic are taken as the input variables to a fuzzy logic knowledge base control system in order to make the admitted control. This scheme also has the property of classifying admission connections to provide differentiated services. We evaluate our scheme with two algorithms: Mosquito algorithm and peak rate allocation CAC scheme. The comparison among these CACs shows our scheme has an improved performance especially in terms of bandwidth utilization. The FCAC also has less computational power than the effective bandwidth method.