AN EFFICIENT FUZZY CONTROLLER BASED TECHNIQUE FOR NETWORK TRAFFIC CLASSIFICATION TO IMPROVE

Abstract

The network traffic management is a key issue today as many new emerging applications are flooding the network with their packets. Several time sensitive and low bandwidth applications run along with other time insensitive and bandwidth intensive applications. As a result, most of the time, channel capacity is exhausted by the bandwidth intensive applications like P2P file sharing, Bittorent etc. Hence, there is limited space left for critical applications. As a result, the accurate identification of network applications through proper observation of associated packets is vital to the areas of network management and surveillance. The overall implication of the accurate traffic management provides space for critical application and also it helps in better management of available network resources like channel capacity etc. The typical approach of traffic classification is based on `well known' TCP or UDP port numbers. This approach is not feasible in current scenario in which applications such as tends to use random port numbers. Other approaches like flow based are able to resolve the issue to some extend but they are also limited to certain protocols or applications. The Payload signature based matching algorithm is able to identify the traffic accurately but again it suffers as the processing time is really high. The work in this dissertation is focused towards exploring the key issues and design factors to be considered for network traffic classification and bandwidth management. An approach is proposed for the network traffic classification and bandwidth consumption by each protocol. It optimizes the usage of the available network capacity and day to day traffic management. The usage of the bandwidth in the networks is managed by taking accurate decisions for various types of applications. The test results show that by using proposed priority based management approach we are able to classify about 90% of traffic accurately. Also the proposed priority based bandwidth management approach is really fast as compare to non priority bandwidth management. The approach is also able to maintain the QoS requirement of the real time connections by marking the time insensitive but bandwidth intensive traffic. These packets can be dropped for proper traffic