5.3 Mobile-Data Network Analytics Highly Reliable Networks

The analysis of mobile network data is a fundamental requirement for the development and invention of novel networking approaches that fulfill the rapidly growing requirements and demands on those networks. This process requires the identification and a thorough investigation of shortcomings in existing field deployments, independent of the network operators and/or the network equipment vendors. Despite public standardization by the 3rd Generation Partnership Project (3GPP), cellular networks are developed and operated as closed systems that provide a predefined networking service to the subscriber while disclosing only a minimum of system-related information such as signal strength or quality in the User Equipment (UE). However, researchers often require a deeper insight into network functionality, especially when it comes to considering network load and occasional congestions while still maintaining the privacy of the regular network users. With this knowledge, future devices may predict their achievable throughput passively under the current load and channel conditions without the need of triggering a transmission just for the sake of throughput measurements that in turn induces (unnecessary) network load. They may leverage this predicted information for e.g. load balancing, network selection, or service adaptation. Since cellular networks are centrally governed by the base stations, which assign the spectral resources by explicit and fine-grained signaling to each active device in the coverage area, information about the cell-wide resource utilization is already “in the air”. For performance reasons and unlike the ciphered payload exchange between UE and the base station, the control messages that carry the resource assignments are not encrypted. However, these messages are scrambled by a device- and session-specific Radio Network Temporary Identifier (RNTI), which is essential for the proper interpretation and validation of those messages and which is exchanged only once at the beginning of each session. This section describes the achievements of the CRC 876 in extracting these control messages of new and already active sessions efficiently and reliably over the air and without the need for expensive specialized hardware. The methodology of the underlying control channel analysis is embedded into a comprehensive opensource software framework Fast Analysis of LTE Control Channels (FALCON), which uses Software-Defined Radios (SDRs) to capture the base station’s signal and accurately extracts the control messages in real time on a regular computer system. In subsequent case studies, supervised learning is used to leverage the disclosed network-load information from short-term observations for the prediction of the expected data rate and ultimately uses this as a metric for dynamic network selection to achieve the highest throughput over the fastest network connection currently available.