TCP Bottleneck Identification in Cellular Network via Passive Monitoring

The bandwidth demand of today’s mobile applications is permanently increasing. This puts particular challenges onto mobile networks where the bandwidth capacity in the radio access network is limited and shared among multiple users. In this work we present a new methodology for detecting capacity bottlenecks in 3G mobile networks by passively analyzing the fine-grained dynamics of TCP flows within a real 3G network. In fact, TCP connections span parts of the mobile network and part of the Internet, and bottlenecks in both network sections may influence the performance. Moreover, since connection states of TCP are maintained only at the two end points (i.e., at the mobile devices and at the Internet servers) we introduce a method for inferring the endpoints’ states at a passive monitoring point in between. By validating our approach based on controlled test traffic in a real 3G network, we show that our method enables us to distinguish different causes of bottlenecks for single TCP flows. The realized monitoring algorithm offers an inexpensive and useful tool for network operators and designers to guide network resource dimensioning and deployment and for on-line performance assessment.