MCNet: Crowdsourcing Wireless Performance Measurements through the Eyes of Mobile Devices

Abstract

Measurement of network performance in complex WiFi networks, such as in corporations and universities, is an important but challenging task, as wireless performance in networks with many access points is hard to measure and model. We demonstrate that crowdsourcing the task of measuring WiFi performance is an effective solution to this problem. By measuring performance directly with unmodified consumer mobile devices such as smartphones, it is possible to cheaply and easily detect problems that matter to users. Aggregated performance data across clients can then provide a global view of performance trends in an enterprise network. We demonstrate that periodic sampling allows the collection of representative data while keeping battery consumption low, and we leverage mobile sensor information to intelligently schedule these measurements. This system was deployed in two different large WLANs, where we discovered numerous previously undetected performance problems.