A Research On The Fault Localization Method For Testing Publish/Subscribe Distributed Systems

The data distribution service based on the publish/subscribe model is a distributed realtime middleware communication specification proposed by the OMG organization.In addition to providing an efficient communication mechanism,it also has abundant Qo S policies.Today,it has been widely used in defense,civil aviation and other fields.However,as publish/subscribe communication middleware is widely used in distributed system,the fault detection can detect faults existing in the current system.It has becoming a major challenge to locate the source of the faults of the distributed system based on the publish/subscribe communication paradigm when the system fails.The features of publish/subscribe middleware and distributed systems mask the fault propagation paths and make it difficult to locate the faults in the system.In this paper,a hierarchical online fault localization method is proposed which can automatically locate the fault when the system detects multiple faults at the middleware level,find out the source of the faults quickly,and help the testers repair the system.Main contributions are listed as follows:1.Proposing a fault propagation model for publish/subscribe distributed systems.The publish/subscribe communication middleware can find out the relationships of faults in the system based on the fault propagation model.The complex relationships between faults are stratfied in a distributed system and presented in different forms.2.Proposing a fault localization method for publish/subscribe distributed systems.When the system detects multiple faults,the publish/subscribe communication middleware can perform fault localization automatically according to the fault localization method and show the fault localization result to the system tester.3.Implementing a prototype system for fault localization method based on the Information Integration Management Software.A few tests are designed for evaluating the system.According to the test results,the prototype system can handle fault localization problems for publish/subscribe systems and the performance of the prototype system meets requirements in actual scenario.