Online Identification of Last-Mile Throughput Bottlenecks on Home Routers
Abstract
We develop a system that runs online on commodity home routers to locate last-mile throughput bottlenecks to the home wireless network or the access ISP. Pinpointing whether the home wireless or the access ISP bottlenecks Internet through-put is valuable for home users who want to better troubleshoot their Internet experience; for access ISPs that receive numerous calls from frustrated home customers; and for informing the debate on regulating the residential broadband market. Developing such a system is challenging because commodity home routers have limited resources. The main contribution of this thesis is to develop a last-mile throughput bottleneck detection algorithm that relies solely on lightweight metrics available in commodity home routers. Our evaluation shows that our system accurately locates last-mile bottlenecks on commodity home routers with little performance degradation.