Multiprocessor Real-Time Scheduling

In this chapter we consider the problem of scheduling real-time applications upon multiprocessors, or equivalently upon multicores, which represents the current trend for embedded and cyber-physical systems. We present first popular models of computation for critical real-time applications. We then present a classification of the scheduler we may find in the middleware of real-time and cyber-physical systems. Those algorithms are compared and their important properties are then analyzed. Two very different kinds of solutions are then presented and compared: the partitioned scheduling and the global scheduling. Important metrics are considered in our comparison like utilization bounds and speedup factors. Lastly, we deal with parallel applications where each program can execute on several processors. Such an intratask parallelism allows to use efficiently multicore platforms.