An approximation analysis of nonpreemptive EDF scheduling

Due to recent advances in multimedia technology, a real-time capability is needed to a certain extent even in information systems, in addition to control systems. In general, information systems are classified as soft real-time systems, which keep operating even if a deadline is missed. In soft real-time systems, it is more important to provide statistical guarantees (probabilistic analysis) than absolute guarantees (deadline guarantees). In current real-time systems, real-time scheduling such as EDF (Earliest Deadline First) is used to meet timing constraints. EDF scheduling simply assigns higher priorities in the order of earlier deadlines. It can be applied to aperiodic tasks without prior knowledge of their starting times. In this paper, it is assumed that each attribute of the task follows a probabilistic distribution. The system performance when a nonpreemptive EDF scheduling is applied is then mathematically analyzed.