PerfMon: Measuring Application-Level Performance in a Large-Scale Campus Wireless Network

WiFi has become one of the most popular ways to access the Internet. However, in large-scale campus wireless networks, it is challenging for net-work administrators to provide optimized access qual-ity without knowledge on fine-grained traffic charac-teristics and real network performance. In this pa-per, we implement PerfMon, a network performance measurement and diagnosis system, which integrates collected multi-source datasets and analysis methods. Based on PerfMon, we first conduct a comprehensive measurement on application-level traffic patterns and behaviors from multiple dimensions in the wireless network of T university (TWLAN), which is one of the largest campus wireless networks. Then we sys-tematically study the application-level network perfor-mance. We observe that the application-level traffic behaviors and performance vary greatly across differ-ent locations and device types. The performance is far from satisfactory in some cases. To diagnose these problems, we distinguish locations and device types, and further locate the most crucial factors that affect the performance. The results of case studies show that the influential factors can effectively characterize per - formance changes and explain for performance degra-dation.