Zoom-inRCL: Root Cause Localization at Virtualized Infrastructure Layer for B5G/6G Network Slicing

Network slicing, as the backbone technique of evolving B5G/6G, consists of Network Function Virtualization Infrastructure (NFVI) layer, network slice instance layer, service instance layer and management and orchestration module. Given the lessons learned from recently reported nationwide and long-lasting global telecommunication disasters with severe service degradation or even outages, we find that many of them orig-inated from a simple faulty entity in the NFVI layer at the very beginning. To the best of our knowledge, as the very first attempt, we propose Zoom-inRCL, which enables us to quickly and accurately find the root cause entity at the NFVI layer once observing the slice service degradation. Specifically, it first filters abnormal NF call graphs at the graph level based on deep support vector data description (Deep SVDD) algorithm and then filters faulty entity candidates at the node level by novelly designed rules, and finally infers the suspicious entity rank list according to the ratio of affected NFs. Evaluations on a real-world dataset show that for over 80% of fault cases, the top-ranked entity identified by Zoom-inRCL is the actual root cause of the service degradation. We believe that our work can provide useful guidance for the design of future B5G/6G networks.