自动化与信息工程2026，Vol.47Issue(3)：42-50,9.DOI:10.12475/aie.20260306
面向分布式系统的无侵入式链路追踪方法研究
Research on Non-intrusive Link Tracing Method for Distributed Systems
曾佳 1王航 1杨晨 1卢晓燕 1张倍嘉 1李豪1
作者信息
- 1. 中国民航信息网络股份有限公司重庆研发中心,重庆 401122
- 折叠
摘要
Abstract
The proliferation of microservices and cloud-native architectures has led to increasingly complex distributed systems,posing significant challenges to observability.Traditional distributed tracing methods often suffer from strong code invasiveness,incomplete link reconstruction,and poor compatibility in heterogeneous environments.To address these issues,this paper proposes a non-invasive distributed tracing method based on multi-source data fusion.This method utilizes the extended Berkeley Packet Filter kernel-level probe to capture network call events and combines them with standardized application logs for context correlation,achieving end-to-end link data collection without modifying business code.Based on the OpenTracing standard,a unified data model is constructed,and a multi-modal storage architecture incorporating relational,time-series,graph,and full-text search databases is adopted to efficiently manage heterogeneous data.Through the collaboration of real-time stream processing and offline batch processing,abnormal patterns such as prohibited calls and cyclic calls are automatically identified.Experimental results show that this method achieves precise reconstruction and visualization of call links while maintaining low resource overhead,significantly improving fault localization efficiency and system operation and maintenance.关键词
链路追踪/eBPF/依赖分析/异常检测Key words
link tracing/eBPF/dependency analysis/anomaly detection分类
信息技术与安全科学引用本文复制引用
曾佳,王航,杨晨,卢晓燕,张倍嘉,李豪..面向分布式系统的无侵入式链路追踪方法研究[J].自动化与信息工程,2026,47(3):42-50,9.
