ClosRCA-Bench: An Open Topology-Grounded Benchmark and Counterfactual Recovery Framework for Self-Healing Datacenter Networks

Read full text Summarise with VIRAT
Authors: Dheeraj Ramasahayam
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Open research on datacenter-network root cause analysis (RCA) is constrained by private production traces and by public studies that omit the topology or remediation context needed for self-healing systems. This paper introduces ClosRCA-Bench, a reproducible topology-grounded benchmark built from Cisco’s public Clos-topology telemetry by combining event files, CDP maps, and per-device YANG telemetry into fixed graph windows. The benchmark contains 311 windows over 11 topology nodes with 30 features per node and four cause families: BFD outage, blackhole, ECMP change, and interface shutdown. Two fault classes localize to hidden target devices that are not directly monitored, making topology-aware localization central. We compare rule-based, correlation-based, graph-free, and spatio-temporal graph RCA methods, and we evaluate remediation through a safety gate and a counterfactual topology digital twin. On the held-out split, Random Forest achieves anomaly F1 0.9688 and cause F1 0.9707, while the full STGNN reaches 0.8380 weighted target-device F1 and 1.0000 hidden-target accuracy; removing topology reduces hidden-target accuracy to 0.0000. These results establish an open benchmark for topology-aware, temporal, and recovery-aware RCA.

Fields of Study & Subjects

Software System Performance and Reliability Software-Defined Networks and 5G Network Security and Intrusion Detection

Publication details

Journal / source
Figshare
Publication date
1 January 2026
Language
EN
Document type
Preprint