| Real-time computing has been subject of numerous research studies in the past three decades or so. Scheduling is probably the most researched area in real-time computing. Fault-tolerant computing, because of its importance in real world applications, has also received a great deal of attention in the past thirty years. Some of these applications require high reliable systems, those that can operate despite failures. In these applications, failure of the system may result in loss of human life and/or financial loss. It is a well established belief in the real-time research community that, to be practical in real life, some of the areas in real-time computing, such as scheduling and fault-tolerance, must be considered jointly and simultaneously. However, many of the research studies in real-time computing have concentrated in one area, independent of the others, or have addressed different areas jointly, but under some unrealistic assumptions. That is, for example, scheduling is either done under the assumption that the underlying environment is fault-free or fault-tolerance scheduling is done under the assumption that there are other mechanisms (software and hardware) that detect the failures instantaneously as they occur.; In this work, we address the problems of scheduling and software and hardware fault-tolerance along with on-line diagnosis of faults in real-time systems simultaneously. We design a wide range of scheduling algorithms for timely execution of real-time tasks in the absence and presence of faults. We evaluate the performance of these algorithms using analytical and/or simulation techniques. To be more precise, this work is divided into three parts. In the first part, we develop optimal (yet to be defined) scheduling algorithms that assume a fault-free environment. The concentration of this study is in the second part, where we assume systems that are subject to hardware and/or software failures. We develop scheduling algorithms that execute real-time tasks in a timely manner, despite failures. One of the developed algorithms not only achieves this goal, but also it detects the processor failures on-line by using application tasks. In this work, we introduce a more realistic fault model than some of the existing works. Finally, we develop a model for evaluating performance of real-time systems that are subject to both static (hardware component) and dynamic (software or timing) failures. This model is more suitable for analyzing real-time systems than some of the existing models, because it directly takes into account those parameters that are of interest in real-time systems. |
