An In-depth Analysis of 3G Traffic and Performance

As the popularity of cellular network grows explosively, it is increasingly important to develop an in-depth understanding of characteristics of cellular traffic and performance. In this paper, we analyze cellular network traces of different generations from three major cities in China during 2010 and 2013 and a city in Southeast Asian country in 2013. We analyze (i) cellular traffic, (ii) throughput, round-trip time (RTT), and loss rate, and (iii) how they are affected by cellular modes, time, and geographic locations. We find the TCP performance in our traces is much worse than those reported in previous 3G measurement studies from North America and Europe likely due to more expensive cellular data plan and more limited cellular resources. We further use machine learning to diagnose reasons behind high RTT and losses, and identify major factors that limit TCP throughput. Our analysis shed light on important characteristics of cellular traffic and performance in large-scale cellular networks.