Research And Implementation Of A Fault-Tolerance Evaluation Approach On The Fault-Tolerant Prototype

As the development of handling of supportive business of typical professions, information services, and application of financial computing for bank and other support businesses, the demand for computing ability of computers is rising rapidly, while the demand for high speed, non-stop and real-time dealing with all kinds of transition processing tasks are also raised. The delay or failure of these computer systems could cause immeasurable loss. In this system, we added an arbitrary processor to guarantee the non-stop running of the whole system, so that it is of great significance to the evaluating of the fault-tolerance performance of this system.The fault-tolerance is usually measured in aspects such as: availability of the system, reliability of the system, and measurability of the system. Availability and reliability are the most important ones among them.In the aspect of availability, we adopt the method of building an availability model for the prototype system to evaluate the availability of the system. Based on the research and analyze of the characteristics and structure of the system, firstly, we build an availability model for the system in ideal condition. We analyze the influence of the failure rate of the components on the performance of the system, and we calculated the availability of the system. Then we built the actual availability model for the system. We analyze the influence that fault-detection coverage and the accuracy of the diagnostic and refracting .we predicted the availability of the system.In the aspect of reliability, we use the fault-injection method to evaluate the reliability of the system. Fault-injection is an effective method to evaluate the fault-tolerance mechanism of the system. We accelerate the process of system failures through artificial faults which are injected to the system. Then we used the results from monitoring and analysing the reaction information after the faults are injected to the system, the performance evaluation of the fault-tolerance system. We have studied the evaluating method of fault- tolerance performance, and we have picked the normal fault types like communication fault, memory fault, register fault, PCI fault as the types of our fault types. Then we design and implement fault injection tools to evaluate the fault tolerance performance of the prototype system.Finally, we get the relative measures of the fault-tolerance performance of the system, and we get two important parameters which are the fault-detection coverage and the accuracy of the diagnostic and refracting under two conditions which are system with arbitrary processor and system without arbitrary processor, then we calculated the availability of the system by using these two parameters, and from that, we concluded that the availability of the system got improved by using the arbitrary processor.