Abstract

A longstanding approach to measuring Internet performance is to directly compare throughput against pre-defined benchmarks (e.g., 25 megabits per second downstream, 3 megabits per second upstream). In this paper, we advocate, develop, and demonstrate a different approach: rather than focusing on whether speeds meet a particular threshold, we develop techniques to determine whether a variety of Internet performance metrics (including throughput, latency, and loss rate) are comparable across geographies.

We define these metrics and apply them across a longitudinal dataset of Internet performance measurements comprising approximately 30 neighborhoods across the City of Chicago. The metrics we define show some geographical disparities, indicating that such comparative metrics may be promising for studying questions of equitable Internet access across neighborhoods.