Kunlun Anomaly Troubleshooter: Enabling Kernel-Level Anomaly Detection and Causal Reasoning for Large Model Distributed Inference

TL;DR

Kunlun Anomaly Troubleshooter (KAT) is a novel framework that detects kernel-level anomalies in large model distributed inference systems with high precision and recall, and provides causal reasoning and natural language explanations to improve troubleshooting efficiency.

Contribution

KAT introduces a new approach combining GPU worker synchronicity analysis with domain-adapted LLMs for precise anomaly detection and causal reasoning in LMDI systems.

Findings

Achieves over 0.884 precision in anomaly detection

Achieves over 0.936 recall in anomaly detection

Significantly improves troubleshooting efficiency and success rate

Abstract

Anomaly troubleshooting for large model distributed inference (LMDI) remains a critical challenge. Resolving anomalies such as inference performance degradation or latency jitter in distributed system demands significant manual efforts from domain experts, resulting in extremely time-consuming diagnosis processes with relatively low accuracy. In this paper, we introduce Kunlun Anomaly Troubleshooter (KAT), the first anomaly troubleshooting framework tailored for LMDI. KAT addresses this problem through two core innovations. First, KAT exploits the synchronicity and consistency of GPU workers, innovatively leverages function trace data to precisely detect kernel-level anomalies and associated hardware components at nanosecond resolution. Second, KAT integrates these detection results into a domain-adapted LLM, delivering systematic causal reasoning and natural language interpretation of complex anomaly symptoms. Evaluations conducted in Alibaba Cloud Service production environment indicate that KAT achieves over 0.884 precision and 0.936 recall in anomaly detection, providing detail anomaly insights that significantly narrow down the diagnostic scope and improve both the efficiency and success rate of troubleshooting.