A fault-tolerant scheduler for distributed real-time systems

Abstract

A distributed algorithm for providing fault-tolerant optimal scheduling in a simply periodic distributed real-time system based on the deadline mechanism is proposed. Each job is serviced by a primary algorithm or by an alternative algorithm. The primary algorithm provides a desirable service that may not satisfy the timing constraints, whereas an alternate algorithm provides acceptable service, while always satisfying the timing constraints. After using the deadline mechanism for scheduling the jobs on individual nodes, the proposed algorithm is invoked to schedule additional primaries on the other nodes. Since primaries provide accurate results, the distributed scheduling algorithm provides better results without disturbing the timing accuracy. The proposed distributed scheduling algorithm was tested for various randomly generated data.<>