On the use of anonymized trace data for performance evaluation in IP routers

Abstract

For IP routers, it is important for realistic performance evaluation of address lookup algorithms to consider both routing table and trace data obtained from target network. However, most of trace data available are published after anonymizing personal information. Thus, the published trace data cannot be applied directly to estimate performance of routers. In this paper, we propose a new method for predicting more realistic router performance by using anonymized trace data. For our motivation, we analyze correlations of address space usage between real trace data and routing table. Based on analytic results, we also propose a method which transforms IP addresses in trace data by using statistics of address space usage in routing table. Through trace-driven simulation we show that our method can predict the routers' performance closer to the actual one. © Springer-Verlag Berlin Heidelberg 2007.