Research And Implementation Of IP Packet Classification Simulation And Evaluation System

While the current Internet offers best-effort service, future IP networks will provide enhanced services to its users. Such services may include differentiated services, fine grain quality of service, etc. All such enhancements need packet classification. However, packet classification on multiple fields is a difficult problem, hence, current algorithms are hardly able to meet the requirments of varied applications and evironments. So the evaluation systems and the functional simulators are needed to help the researchers simulating and evaluating current algorithms and new ones. With the valuation systems and the simulators they can find out the better algorithms for the specified situation, or can test their state-of-the-art ideas of the new algorithms. Also the developers of the network equipments related to packet classification can find valuable advices from the evaluations.After analyzing the characteristic of the packet classification problem and studying the performances which researchers and developers are interested, we define several metrics to help evaluate the packet classification on a fine scale. Based on the consideration on the metrics and packet classification environments, we propose a general performance evaluation system model of packet classification and try to standardize the method of performance evaluation. According to our theory analyzing on the algorithms and the evaluation system model, quantity of improvements are done on the Packet Lookup And Classification Simulator(PALAC). These improvements involve new Filters Generation Module, new Hardware Algorithm Supportive Interface, the refinement on the Statistics Collection and Query(STAT) Module and the console. This performance evaluation system provides a better tool to evaluate classification algorithms. At the end of this paper, some typical classification algorithm is tested with the evaluation system, and some conclusions are drawed. This may provide some advices for the future research on packet classification.