A Flexible Framework to Schedule Soft Aperiodic Tasks in Hard Real-Time Systems

An EDF-based flexible scheduling framework designed for scheduling multiprocessor real-time systems composed of soft aperiodic and hard sporadic tasks is described. Hard tasks are assigned to processors and do not migrate between processors during their execution. Migration of soft tasks is allowed. They can be put either in a global queue, accessed by all processors, or in each processor local queue. We show that using the framework in either configuration provides improvements in QoS-related metrics for soft tasks, such as deadline miss ratio and response time without compromising timeliness guarantees for hard tasks. In the core of our solution is the use of two types of servers, whose definition and correctness proofs are also presented in this paper. Our solution is evaluated via extensive simulation. Results indicate that the available processing capacity left unused by hard tasks can be effectively used for executing soft aperiodic tasks.