Dependency Tree Based Adaptive Error Detection And Error Diagnosis

This paper focuses on error detection and diagnosis in distributed networks or systems, with considerations of current problems of error detection and error diagnosis, proposes an approach to deriving a dependency tree from a dependency matrix, and develops a novel dependency-tree based adaptive probing method for error detection and diagnosis. A dependency tree, as we explained in the paper, can be used to organize probes for the process of probing, and provide much more information about the relationship between destination system component and probes, also the one among different probes, which is the specific characterization of dependency tree. Therefore, it can be the basis of probes set selection according to the system status for error detection and diagnosis. The proposed dependency-tree based probing method, which results in adaptive and optimal probing, can significantly reduce network traffic load associated with fault diagnosis owing the decrease of probe sets selected through the algorithm, also can greatly reduce the size of probe sets, and greatly remove the overload of network.The paper begins with the core algorithms, one is the algorithm of transformation from dependency matrix to dependency tree, and another dependency tree based adaptive probe. And the description of how to solve the error detection and error diagnosis is also included. Moreover, the paper extend the adaptive algorithm to include the integrate probe framework in order to solve the data sharing and other problems such as the communications between different probe agents and to detailed describe the needed components in the framework and its character and specification. Finally, the paper, using the experiment data collected by experiment, detailed describes the efficiency comparison in the course of error detection and error diagnosis between the dependency tree based algorithm and other ones. Through the experiment data comparison and analysis, it is easy to know that the algorithm the paper used can keep the success rate of error detection and error diagnosis using the selected probe set, which brings the minimal overload for networks or systems.At last the paper give a summary of the algorithm and framework, the innovation and further discussion of dependency tree and its based method for error detection and error diagnosis.