Font Size: a A A

Research On Diagnosis Technologies Of Distributed Component System

Posted on:2006-01-29Degree:DoctorType:Dissertation
Country:ChinaCandidate:J HuangFull Text:PDF
GTID:1118360185463790Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Widely-spread component technologies provide the seamless integration of infrastructures, the reusability of binary code and the capability of rapid application development for distributed applications in past twenty years. While the complexity of developing component-based distributed applications is reduced greatly, there are still difficulties in debugging and diagnosis. Thus, diagnosing and locating of faults efficiently inside the large-scale distributed systems are imperative for reducing mean time to repair and keeping the high reliability and usability.This dissertation aims to construct diagnostic supporting systems on the distributed component platform. The software build by the distributed component technologies is a complicated system. Compared with hardware, software is more sophisticated and has special diagnostic goals. By the analysis of existing fault diagnosis systems and methods, this dissertation proposes the model-based diagnosis solution for semantic diagnosis problems of distributed components. With conceptual classification of the current formalized diagnosis models, we target the consistency-based diagnosis algorithm and improve it at multiple aspects. Such as, the algorithm to resolve minimal over-determined subsystem, the algorithm to calculate conflicts and the algorithm to compute minimal diagnosis, etc. We study the temporal features of semantic diagnosis in distributed component systems, and then provide a diagnosis model with temporal characteristics. Special for memory faults diagnosis, we present the BMMD method based on Bayesian theory and the criteria of diagnostic features selection. Finally, we studied the implementation mechanisms of semantic diagnosis for assembly and memory faults diagnosis on the basis of StarCCM platform, which make StarCCM platform easy to use.The main contents of this dissertation are as follows:1. We investigate the challenges to diagnosis on distributed component systems, and then focus the research works on providing diagnostic mechanisms for component based applications. After analyzing the traditional diagnosis theories as well as the current distributed component fault diagnosis systems, we present three conceptual diagnostic models according to system information and observations used by diagnostic methods. They are deviation from normal behavior (DNB) diagnosis, matching abnormal behavior (MAB) diagnosis and abnormal classification (AC) diagnosis. Then we compare them in diagnostic knowledge, formalization methods and reasoning methods. According to three conceptual models, we analyze the consistency-based diagnosis, abductive diagnosis and hypothetical diagnosis.2. In order to improve the efficiency of diagnosis, we present an algorithm named SAMODS which computes the minimal over-determined systems. This algorithm takes full advantage of redundant information during diagnosis and decreases the calculation in real diagnosis stage by means of computing all minimal over-determined systems statically. The conflict recognition algorithm, named MODSBCI, making use of the SAMODS algorithm results is also provided. This dissertation presents a compound simulated annealing and genetic algorithm, named GSA, to solve the minimal diagnosis by mapping the problem to 0/1 integer...
Keywords/Search Tags:Distributed Computing, Model-based Diagnosis, CORBA Component Model, Minimal Diagnosis, Bayesian Method
PDF Full Text Request
Related items