Mobile communication is often frustratingly slow. When a user encounters poor performance, and perhaps even “confirms” the same by running a speed test, the tendency is to ascribe blame to the user’s last-mile provider. However, as we argue in this paper, a more nuanced approach is needed to identify the location of the bottleneck responsible for the poor performance. Specifically, we focus on the question of whether the bottleneck lies in the cellular last hop (3G or LTE link) or elsewhere in the WAN path.
We present QProbe, a tool that takes advantage of the proportional fair (PF) scheduler employed in cellular networks to determine whether queuing is happening at the cellular link. After validating QProbe through simulations and controlled experiments, we present our findings from a measurement study conducted over a 2 month period involving over 600 real-world users across 51 operator networks in 33 countries. We find that, for example, the cellular last-hop link is the bottleneck in 68.9% and 25.7% of the total bottleneck cases for 3G and LTE clients, respectively, suggesting that there is a significant fraction of cases where the poor performance experienced by the user is due to the WAN and could potentially be routed around. Moreover, we show that QProbe detects the bottleneck link location with greater than 85% accuracy for both 3G and LTE clients in our measurement study.
Following are the results of QProbe test, on an iPhone 6, taken at my home in Tel-Aviv (Saturday morning, in a quiet neighborhood, is probably not the best time to diagnose bad connections):
See "QProbe: Locating the Bottleneck in Cellular Communication" - here.