Proactive Probing and Probing On Demand in Service Fault Localization

摘要

This paper proposes Proactive Probing and Probing on Demand (pPOD) approach to monitor the health of a serviceable component system (SCS) proactively and to localize faulty serviceable component on demand. In general, many highlevel web services have dependency on other web services and on the basic network infrastructure. The health of such high-level services relies on the health of all these services and the network infrastructure. Therefore, we can select a small set of such high level web services and monitor them by probing their health periodically. This application-level performance monitoring mechanism can greatly reduce the overhead of satisfying servicelevel objectives. When abnormal events occur, we are able to follow the dependency path to localize possible failures based on the outcomes of incremental on demand probing. The proposed approach effectively reduces the resource and traffic cost in performance monitoring and speeds up the diagnosis process when compared with traditional performance monitoring and fault localization methods.