Analysis of Selective Fault - Tolerant , Hard Real - Time

An increasing number of applications are demanding real-time performance from their multiprocessor systems. For many of these applications, a failure may produce disastrous results. Such failures are avoided in hard real-time systems by the use of fault-tolerance. In hard real-time multiprocessor scheduling, this fault tolerance may be provided by including several task backups in each schedule. Selective fault-tolerance is the practice of providing a higher level of reliability for critical than for nonessential tasks. Multiprocessor schedules are said to be selectively fault-tolerant if they contain backups for critical tasks, thus giving critical tasks a higher reliability. The area of fault-tolerant hard real-time multiprocessor scheduling of critical and noncritical tasks has not been carefully studied. We have derived upper bounds on the length of selectively fault-tolerant, hard real-time schedules produced by several algorithms. In addition, we have developed algorithms that build selective fault-tolerance into existing hard real-time multiprocessor schedules. Our experimental results reveal the weakness of relying on one type of fault-tolerance, and the strength of combining several types in an e ective way. Furthermore, these results provide valuable information which might aid in creating an even more e ective way of combining fault-tolerant methods.