Research Of Software-Implemented Fault Injection And Reliability Evaluation Methods In Distributed Real-Time System

The computer system gets more and more wide application in some high reliability fields such as aerospace, nuke industry, traffic control, banking and medical etc. In these fields, failure of computer system may cause enormous personnel casualties and economic loss. So how to test and verify validity and availability of the fault tolerance system before applying them for making sure they are consistent with expecting, it needs effective theories about evaluation of fault tolerance. Reliability verification of computer system is a complicated endeavor, normally using modeling theory and method. But due to the fault activation mechanisms and error propagation process is very complicated so that they can't be completely known, it is hard to modeling the real system. Fault injection is artificially producing fault and introducing to the target system accelerating system errors and failure, then by analyzing system reaction to acquire reliability and fault tolerance evaluation results of target system. Compared with other methods, fault injection has brought the scholars'attention because it has an advantage in wide application, more accurate result and less time spending.Currently, distributed systems and distributed real-time systems are widely used in a large number of key industries. There is great difficulty and high demand on the reliability assessment and measurement of these systems and the trusted monitor of the system as a whole is difficult. According to this situation, this paper has an in-depth study of fault injection technology. It is explored software-implemented fault injection methods for distributed real-time systems and established corresponding fault injection architectures. Taking into account that the distributed network environment is also an important factor that can impact the dependability of real-time system, we also have an in-depth study of the network fault detection and diagnosis. In this paper, it is researched on the theory and technology of reliability models and evaluation methods and explored evaluation methods of reliability index. Specific studies are as follows:①Researching the features and contents of software reliability, especially the differences and meaning contacts between software reliability and fault injection, the transformed relationship among defect, error and failure, and the characteristics and architectures of distributed real-time systems; This paper will be to provide background knowledge for the further research of software-implemented fault injection, reliability model establishment and evaluation technology in module and system level.②Discussing and studying the fault models of distributed real-time systems, directing the problem of dependable verification, analyzing the features of fault injection of distributed real-time system, establishing a common model of the fault injection. On the basis, it is summarized the ways of transcacting fault and the classification of controlling time of the fault model.③This paper researchs and analyzes the method and frame of fault injection, proposes a software-implemented fault injection system structure based on distributed real-time system, which are divided into three levels: the target system, the communication network, the software-implemented fault injection system. It is designed software-implemented fault injection system, which contains software fault injector, data collection modules and fault data analysis module. We further research into software failure on the bottom of the principle and realization mechanism and put forward the executive algorithm of fault injection.④In this paper, it is researched and analyzed the basic concepts of the theory of evidence and model, proposed a method of network fault detection and diagnosis based on the theory of evidence, explored the effective and veracious degree of improving fault diagnosis based on the theory of evidence, analyzed and constructed a network reliability classification index model.⑤According to the efficiency and veracity limitations of the present software reliability models and the characteristics of real-time executive of distributed real-time system, it is set up a reliability model which can better reflect distributed real-time system characteristics. It is summarized the basic principles and the reliability design method, and put forward the method to determine the key link and the method to determine the reliability of the system in different conditions and the method to determine the most reliable communication path set. It is proposed a method of reliability assessment based on the time limit and gave a FST algorithm related with program and the evaluation tidy algorithm of responsing time to reduce the system time costs; The reliability evaluation methods of distributed real-time system is also proposed. It is analyzed the quantitative relationship between the system reliability and software module and recognized the software modules which to influences more to the system reliability.