Analytical performance evaluation of Internet access over GPRS and its comparison with simulation results

We discuss the results of recently published analytical concepts for the performance evaluation of packet-switched mobile radio networks (see Vornefeld, U., Proc. European Wireless 2002, 2002; Proc. Int. Commun. Conf. ´02, 2002). We compare them with the results of a simulation system (GPRSim), which emulates the GSM general packet radio service (GPRS) in great detail (see Stuckmann, P., Technical Report, http://www.comnets.rwith-aachen.de/∼pst; Stuckmann and Paul, O., Proc. 10th Symp. Wireless Personal Multimedia Commun., p.597-602, 2001). We try to identify possible reasons for deviations between analytical and simulation results in order to support further development in this area. In Vornefeld´s analytical modelling framework, source modelling is done by characterization of WWW traffic by means of marked Markovian arrival processes (MMAP); accordingly, GPRS performance measures are derived by analysis of an MMAP/G/1 queue. By evaluation of WWW traffic generated during stochastic simulation, we show that important characteristics of recent Internet traffic are depicted by Vornefeld´s analytical source model. We find that the analytical approach is well suited for understanding the general impact of various system parameters. Nevertheless, although the analytical model uses a very detailed characterization of WWW traffic, for quantitative performance analysis, simulation results are still required.