Design And Implementation Of Fault Injectors For High-End Fault-Tolerant Computer

High-end fault-tolerant computer means computer that has both high performance and high availability. High-end fault-tolerant computers are widely used in financial, telecom field and other fields, computer systems used in these crisis fields require extremely high reliability and availability.In order to provide high availability, many fault-tolerant methods including hardware layer, firmware layer and software layer, are used in high-end fault-tolerant computer. Fault-tolerant capability evaluation is necessary in the research and development process of high-end fault-tolerant computer.Fault injection is the most common and most effective method to evaluating the fault-tolerant capability. This paper focus on how to use fault injection to evaluate fault-tolerant capability. On one hand fault injection model is studied, on the other hand fault inject tools are designed and implemented.Based on FARM fault injection model theory, this paper constructs a fault injection model aimed at high-end fault-tolerant computers. The faults set F, activations set A, readouts set R and derived measures set M are defined to describe the strategies of fault injection.According to the fault injection model, 9 fault injection tools need to be implemented in application layer, system software layer, firmware layer and hardware layer. This paper is in charge of 3 fault injectors: ptrace based register fault injector in application layer, physical address memory fault injector in kernel layer and network device fault injector in device driver layer.In order to validate the effectiveness of designed fault injections, and to explain how to evaluate fault-tolerant capability, this paper selects several high-end fault-tolerant Itanium architecture computer to perform fault injection experiments. When fault injection is considered, the input domain corresponds to F and A sets and the output domain corresponds to R and M sets.The fault-tolerant capability of target system is analyzed from the fault injection results. It's proved that our fault injection model and fault injectors are effective to evaluate the fault-tolerant ability of high-end fault-tolerant computer systems.