Tackling Bufferbloat in 3G/4G Mobile Networks

With the exponential growth of hand-held devices like smart phones and tablet computers, a deep investigation of TCP performance in cellular networks is becoming increasingly important. In this paper, we conducted extensive measurements over the 3G/4G networks of four major U.S. carriers as well as the largest carrier in Korea and discovered a significant problem: the bufferbloat in cellular networks nullifies loss-based congestion control and allows excessive growth of the TCP congestion window, resulting in extremely long delays and throughput degradation. To alleviate this issue, smart phone vendors put an upper bound on the sender’s congestion window by advertising a static receive window smaller than the actual receive buffer size. This simple trick helps mitigate the problem, but has fundamental limitations due to its static nature. In this paper, we propose a dynamic receive window adjustment (DRWA) algorithm to improve TCP performance over bufferbloated cellular networks. According to our extensive real-world tests, DRWA may reduce the delay by 25 ∼ 49% in general cases and increase TCP throughput by up to 51% in some specific scenarios. Since DRWA requires modification at the client side (e.g., smart phones) only and is fully compatible with existing TCP protocol, it is immediately deployable.