Practical Performance Degradation Mitigation Solution Using Anomaly Detection for Carrier-Grade Software Networks

Abstract

One of the critical technologies required for the large scale acceptance of network functions virtualization within carrier-grade communication systems is the maintenance of a predictable performance level for the software network functions. However, due to the specific resource allocation and to the implementation of the software itself, the performance of the network functions tends to degrade in time, thus, reducing the capacity of the system to serve the specific service requirements. This article introduces a practical solution for performance degradation detection and mitigation for telecom oriented software networks. Also. It includes the mechanisms to interact with the software network and a mechanism for abnormal behaviour detection, as basis for the performance degradation automatic decisions. Furthermore, the solution is exemplified as an addition to the current 5G core network architecture and evaluated on a testbed based on the Fraunhofer FOKUS Open5GCore toolkit and IBM DeepAD. The measured results show that such a solution is feasible and should be further investigated to be integrated in the next generation carrier-grade software networks management for enabling an autonomous long duration functioning of the network.