Dynamic feedback and elastic scheduling model for flexible workload management

Soft real-time systems, such as multimedia applications, are becoming increasingly unpredictable due to the variable of the tasks' attributes. Those scheduling algorithms, depending on the tasks' static attributes, can't allocate system resource efficiently. In this paper, we present a dynamic feedback and elastic scheduling model and period adjustment algorithm for flexible workload management. Instead of using tasks' resource utilization, our algorithm uses the number of missing deadline as the parameter for period adjustment. Our scheduling model samples the system state, and adjusts the tasks' period to change the tasks' resource requirement in the next sampling period. In this way, our model make the system converge to the steady state when the system is overload, and use system resource utilization sufficiently when system resource isn't fully utilized. We analysis the model and evaluate its performance.