Component analysis of GPRS traffic based on user-level from a live network

Characterizing aggregated GPRS traffic and user traffic is important for performance evaluation and optimization of the networks. In this paper, we investigated these characteristics and their relationship. Our analysis are based upon user-level flow traces from passive monitoring of more than 1000 GPRS/EGPRS cell sites connected to the same SGSN at Gb interface. The traffic traces are from a major China ISP in one urban area of Northern Province during an ongoing 24 h period in August 2012. We found that user traffic volume in a 5-minute period at a given cell is consistent with a heavy-tailed distribution. A small portion of large volume users accounted for more than 70% of downlink resource of cell. Furthermore, with the increase of arriving users in a 5-minute period at a cell, the large-volume users contribute more traffic proportion of total cell traffic. These findings are useful for constructing a practical and predictable mobile internet traffic model.