A novel dynamical approach to (m, k)-fırm scheduling
Milton Armando Cunguara, Tomas Antonio Mendes Oliveira e Silva, Paulo Pedreiras · 2014
Many real-time systems tolerate, up to some extent, that instances of their tasks are skipped or miss their deadlines. In such cases it is possible to increase the system schedulability by dropping some jobs. However, the number and temporal distribution of such drops must be constrained, in order to bound its impact on the system performance. The (m, k)-fırm model, which specifies that at least m out of any k consecutive jobs are scheduled, captures such constraints and has been a topic of interest of the scientific community over the last few years. Many advances in the implementation of this type of schedulers were made. However, the (m, k)-fırm schedulers hitherto proposed in literature either do not provide deterministic (m, k)-fırm guarantees or are not efficiently scheduled. In this paper it is presented a (m, k)-fırm scheduler that provides deterministic (m, k)-fırm guarantees and is capable of scheduling task sets that are deemed unschedulable by current approaches. The proposed scheduler works by dynamically changing the underlying (m, k)-fırm frames. Examples of task sets that are schedulable by the scheduler proposed in this paper but that not schedulable by any other scheduler described in the literature are presented.