Reconfigurable Resource Scheduling with Variable Delay Bounds

Certain emerging network applications involve dynamically allocating shared resources to a variety of services to provide QoS guarantees for each service. Motivated by such applications, we address the following online scheduling problem belonging to the recently introduced class of reconfigurable resource scheduling problems: unit jobs of different categories arrive over time and need to be completed within category-specific delay bounds, or else they are dropped at a unit drop cost; processors can be reconfigured to process jobs of a certain category at a fixed reconfiguration cost; the goal is to minimize the total cost. We study this problem in the framework of competitive analysis. Through a novel combination of the EDF and LRU scheduling principles, we obtain an online algorithm that is constant competitive when given a constant factor resource advantage over an optimal offline algorithm.