Self-Organizing Temporal-Spatial Failures Awareness for Access Nodes under Dynamic Mobile Edge Networks

To enable the efficient management and maintenance of the mobile edge networks with growing complexity, an in-time failures awareness is necessary. Besides, the wireless connection between access node and user equipment (UE) is much more vulnerable compare with the physical link. 3GPP has proposed self-organizing networks (SON) but isn’t completely suitable for the mobile edge network. For the classical access node represented by base stations (BS), this paper refers the conception of SON and proposes a two stages self-organizing temporal-spatial failures awareness algorithm with enhanced efficiency. First, with BS-side data such as connected user number, transmit power, and throughput, STL decomposition algorithm and GESD outlier detection algorithm are used to initially detect the potential failure node set. For each BS in the detected potential failure BSs set, multiple UE-side data such as SINR, signal strength (SS), and neighboring access node signal strength (NSS), are further used. Their temporal distribution difference degrees are transformed into a spatial pixel map centered on the BS. Then, CNN network is taken to judge whether the BS is under outage. By simulating under the scenario with 16 BSs, our algorithm achieves 97% recall rate. At the same time, the average execution time of the algorithm is reduced by 86% compared with the KNN-based algorithm.