A new dynamic scheduling algorithm for real-time homogenous multiprocessor systems

Many time-critical applications require predictable performance and tasks in these applications have deadlines to be met. In this paper, we develop a new algorithm for non-preemptive scheduling of dynamically arriving real-time tasks (aperiodic task) in multiprocessor systems. The new algorithm includes two policies: task selection policy and processor selection policy, named grouping policy and properly choosing policy (PCP), respectively. Grouping policy and PCP aim to maximize the effective utilization of resources and processors, respectively. Simulation results show this algorithm outperforms the myopic algorithm and the thrift algorithm.