IMPROVING APPLICATION LAYER IP QOS USING FUZZY LOGIC

Abstract

Mostly, IP QoS solutions are developed from the entire network point of view i.e. developers tried to cover all the aspects of a huge network like scalability etc. Unfortunately, due to the cost and the complexity of such solutions, still the besteffort service is the main working solution. Our aim is to develop a performance diagnostic solution that continuously observes the performance of the network. Whenever the performance degrades to some threshold level, system automatically tries to resolve the performance bottleneck by using QoS principles and mechanisms. For performance measurement we used passive measurement using polling as data extraction technique. There are two major problems in this regard; • How we can measure the available capacity because due to sharing of common bandwidth it is constantly varying? • How can we reduce the number of measurement samples needed to construct the network history? To address first problems our solution uses fuzzy logic, because due to the complexity of solution it is difficult to establish precise values. Hence, we generalized some fuzzy rules and based upon this fuzzy rule base we tried to attain peak performance. To address the second problem, adaptive sampling is used, so when certain predetermined conditions are met, such as an increase in drop rate, corresponding actions are taken such as a decrease in sampling interval and vice versa.