Zoom-inRCL: Fine-grained root cause localization for B5G/6G network slicing

Network slicing, a cornerstone technology for the evolving B5G/6G, comprises the Network Function Virtualization Infrastructure (NFVI) layer, the network slice instance layer, the service instance layer and the management and orchestration module. Reflecting on the lessons from recent large-scale and enduring telecommunication disasters, which resulted in severe service degradation or even complete outages, we observe that many of these incidents stem from an initial simple malfunction within the NFVI layer. To the best of our knowledge, we are the first to investigate the fine-grained Root Cause Localization (RCL) issue for B5G/6G network slicing deep into the NFVI layer, with our proposed Zoom-inRCL scheme capable of not only localizing the malfunctioning entity but also identifying the fault-related metrics. Specifically, it first utilizes the Deep Support Vector Data Description (DeepSVDD) algorithm to filter abnormal NF call graphs from NF invocation traces collected during periods of service degradation in slices, and then identifies suspicious faulty entities through a uniquely designed scoring method, and finally ranks metrics exclusively for those entities exhibiting high faulty scores. Performance evaluation conducted on a real-world mobile operator dataset demonstrates that Zoom-inRCL, by progressively filtering out unrelated noisy data, outperforms existing schemes in RCL accuracy while simultaneously maintaining low time costs. We believe our design idea can enhance the assurance of intelligent and effective operation in the future B5G/6G network slicing.