A hybrid approach to quality measurements in mobile networks

The mobile-first era is here. Due to constantly increasing demand for mobile data, cellular network infrastructures running on limited radio spectrum are struggling to keep up. Constant monitoring and measurements are necessary to ensure service availability and quality as saturated networks are not able to deliver consistent experience. Such measurements are especially crucial for mission-critical use cases such as public safety as they increasingly rely on commercial networks. However, measuring mobile networks at scale bears some fundamental problems. Although there are significant improvements in capabilities of mobile networks (e.g. bit rate, latency), measuring them is still rather complicated task compared to fixed networks given that in mobile networks, performance is a result of complex interaction between momentary cell load, adjacent cell interference, shadowing, fading, mobility and user device capabilities. The adoption of commercial 5G networks is expected to increase the variability even more as it depends on smaller cells. Active measurements that inject large amounts of traffic into the network for the sole purpose of measuring are costly in terms of both bandwidth and energy. Passive mechanisms are lightweight but miss the information of why a certain bit rate is received or sent by the end device. They can not tell whether the performance bottleneck is in the network or in the service itself. On the other hand, modern mobile platforms provide a great diversity of networked applications that have varying network resource demands. It is often hard to estimate which aspect of mobile network performance is relevant to the quality that users experience. For example, just providing an excess of downlink bandwidth at the expense of packet loss or latency might not cut it for highly interactive real-time applications. By combining active and passive measurements in a novel way, this work focuses on a hybrid measurement approach to quality measurements in mobile networks. First, an efficient and scalable hybrid methodology is proposed and evaluated for mobile network Quality-of-Service measurements. Then building on top of it, mobile Quality-of-Experience and its predictability in the field via smartphones is empirically investigated by carrying out extensive field studies. The thesis concludes by evaluating the findings to establish future work necessary to achieve pervasive mobile measurements that are capable of both reflecting user-perceived quality and predicting mobile performance as an enabler of network adaptive applications.