Performance Prediction Method for Address Lookup Algorithms based on Statistical Traffic Analysis

Many address lookup methods for IP routers to improve a packet forwarding capability have been recently proposed, but evaluation on their performance prediction is very limited because of lack of consideration on actual traffic characteristics. It is necessary to consider the actual traffic characteristics to predict more realistic performances of routers, and in this thesis, we propose methods for predicting the router’s performance based on the statistical analysis of the Internet traffic. Our method consists of three steps: (1) creates destination addresses based on traffic statistics derived from traced data, (2) models the tendency of packet arrivals by using ISGF, whose parameter is determined by analyzing the actual traffic, and generates address sequences by arranging addresses created in step (1), and (3) implements address lookup algorithms on a trace-driven simulator and evaluates their performance using traffic patterns generated in step (2). From simulation results, we show that our method can provide accurate performance prediction for address lookup algorithms. Furthermore, we use an approximate mathematical approach for quick performance prediction and discuss its accuracy compared with simulation results.