Resource Freshness in Real Time Scheduling

We define and study a real-time producers/consumers problem, where the production/consumption relations are nondeterministic. In other words, a consumer can get one unit of resources from any one of the producers, and one producer can provide one unit of resources for any one of the consumers. Furthermore, a producer for one type of resourcesmight be the consumer for another type of resources. A class of fresh resourceallocation policies is developed not only tomake resources available for periodic real-time tasks but also to control the resource freshness. We first model the real-time scheduling problem as a feasible minimum cost flow (MCF) problem. Then, by defining the linear cost function differently, we get resource allocation policies for making trade-offs between resource freshness and task schedulability. Finally, we show that our result can be used to extend the priority ceiling protocol (PCP) to work with general unbounded semaphores, whereWait and Signal operations need not be in pairs in one task.