I. INTRODUCTION
THE QUALITY of service (QoS) of mobile applications such as mobile cloud gaming or live video streaming relies on certain bandwidth and round-trip time (RTT) guar- antees [1], [2]. Also the profit of commercial web services depends on low network latency [3]. To optimize QoS accord- ing to available network capacity and avoid failing connections such as dropped VoIP calls, mobile applications need to rely on information about the performance of the underlying mobile network. This depends on a number of aspects, including the number of active users, their location, time, and other factors. One approach improving mobile applications’ performance is to use information provided by mobile operators. However, performance figures advertised by mobile operators in many cases report primarily maximum values that often not match the actual performance [4]. Moreover, detailed information about users per cell, backbone capacity and also expected delays and available throughput are not available at the required
granularity for confidentiality reasons.