Active Diagnosis of High-Level Faults in Distributed Internet Services

摘要：

For fault Diagnosis in internet service,the detection and localization of high-level failure is very important and a real big challenge.The diagnose methods that passively collect information have two drawbacks:1) requiring the target system to report its inner message;2) its impossible to detect and locate faults before user senses them.This paper proposes an active diagnose method which test internet service with probes and make fault inferences based on the probe results.Probing method is proactive and adaptive with low cost.We evaluate it through applying it to a J2EE application Pet Store,compare it with a current passive method Pinpoint,and show that our method outperforms Pinpoint.

关键词： Fault Diagnosis High-level Fault Dependency Matrix Extended Dependency Matrix Probe.

作者: Huihu Long Lu Cheng Yongguo Zeng Li Wu

作者单位: State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications,China,100876

会议类型: 国际会议

会议名称: 11th Asia-Pacific Network Operations and Management Symposium(APNOMS 2008)(第十一届亚太网络运行和管理国际研讨会)

会议地点: 北京

会议语种:英文

页码: 72-81

在线出版日期: 2008-10-22（万方平台首次上网日期，不代表论文的发表时间）