A measurement based solution for service quality assurance in operational GPRS networks

Measuring data and signaling traffic in an operational network may prove to be a better basis for dimensioning than earlier approaches based on simulation models or analytical calculations. Dimensioning rules based on an ongoing measurement process are expected to adapt automatically to changes in traffic statistics and network parameter settings. In an operational cellular network cell load shows diurnal as well as spatial variations. If we look at different cells over time or at cells located in distant geographical areas, we find cells serving varying numbers of subscribers and carrying varying amounts of circuit switched and packet switched traffic. The idea is to capture all user transactions over a large geographical area, determine the traffic load during these user transactions at cell level, and calculate the quality of service measures of interest for the transactions. If samples are collected at different traffic load levels, and we depict the obtained quality of service measures as a function of the traffic load, the resulting graph gives a dimensioning curve. We present a methodology for obtaining cell dimensioning charts based solely on traffic traces available in all operational GPRS networks, and we provide new insights into the performance of the GPRS system by analyzing a month long measurement trace captured in a live network using the new methodology.