Scholarlink

In a distributed microservice system, the complex service instance invocation relationships and the rapid propagation of service failures along the invocation chain pose significant challenges for identifying the root causes of failures. By integrating graph convolutional networks (GCN), we introduce an advanced microservice fault root cause localization model (R-FCLM3). Initially, the model employs eBPF technology to gather observability data, constructs a service call directed graph, extracts node and link features, and utilizes GCN to detect abnormal service nodes and links, while evaluating the severity of each abnormal node. Ultimately, we validated the performance of the proposed R-FCLM3 model on the TrainTicket and Sockshop datasets, and the experimental results demonstrated that the model surpassed existing baseline models in terms of fault localization accuracy and speed.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Microservice Fault Root Cause Location Model Combined with Graph Convolutional Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yang Chen,
  • Zhao Chunxi,
  • Tong Bo

摘要

In a distributed microservice system, the complex service instance invocation relationships and the rapid propagation of service failures along the invocation chain pose significant challenges for identifying the root causes of failures. By integrating graph convolutional networks (GCN), we introduce an advanced microservice fault root cause localization model (R-FCLM3). Initially, the model employs eBPF technology to gather observability data, constructs a service call directed graph, extracts node and link features, and utilizes GCN to detect abnormal service nodes and links, while evaluating the severity of each abnormal node. Ultimately, we validated the performance of the proposed R-FCLM3 model on the TrainTicket and Sockshop datasets, and the experimental results demonstrated that the model surpassed existing baseline models in terms of fault localization accuracy and speed.