Optimized Injecting Method of Software-Implemented Fault Injection Technique

As an analysis method of software dependability, software-implemented fault injection (SWIFI) has been concerned and studied for a long time. Many injectors have been designed to implement injecting experiments. These injectors run on different platforms and have different test goals. In this paper, a software-implemented fault injector is designed to run on Windows NT platform. The aim is to test and evaluate reliability of software that will work in high-radiation environment. In such environment, SEU (single event upsets) is a main reason that causes failure of software. In our injector, SEU is emulated by flipping one-bit or multi-bits in memory or registers and memory injecting is the research focus. In order to perfect injecting theory and experimental technique, some related issues are studied in this paper. Based on memory space injecting technique, two injecting methods are designed and time complexity is computed respectively. Two kinds of injecting experiments are implemented according to these methods. Usually fault injecting experiment is time-consuming and ineffective, so it is important to choose an injecting plan that can optimize tests. The different distributions of location random variable X are used to implement experiments. Through comparative analysis of the experimental results, a conclusion is made that there is always an injecting plan that can significantly optimize experiments.