Surge Handling as a Measure of Real-Time System Dependability

Traditional reliability measures for computer systems can be classi ed into Computer-Centric or Application-Centric categories. The former concentrate on the hardware resources while ignoring the application's needs. The latter focus on the requirements of a speci c application which is being executed, thus requiring the knowledge of all the details of the application; information which may not always be readily available. Also, the narrow view on the system's reliability through a single application is too restrictive and provides very limited information regarding the way the system will handle other applications. In this paper we present new measures for real-time system reliability. These measures are application-sensitive rather than application-centric, and are especially suitable for systems executing various applications with di erent attributes, some of which may not be known in advance. Our proposed measures capture the capability of a real-time system to respond successfully to unexpected surges in the workload. These surges may result from a phase change in the system's mission, an application-related emergency situation or the failure of some system resources. The ability of the system to handle such surges determines, to a large extent, its chances of survival and meeting its applications' deadlines.