Fuzzy constraints based predictive utilization control for soft real-time systems

To guarantee performance of soft real-time applications operating in unpredictable environments, a predictive control based feedback scheduling is proposed. The performance metrics, utilization and tasks deadline miss ratio, are both considered in the cost function, while the corresponding weight matrixes are decided by system workload state. The control objectives are defined by their satisfaction degree represented with fuzzy membership functions, thus the control efforts achieves a balance between two metrics in the sense both with highest satisfaction degree when system overloaded. The optimization problem is solved by converting it into quadric programming problem subject to soft constraints. Simulation results show the proposed approach could achieve favorable performance and is robust against workload uncertainties.