Delayed Rate Monotonic Algorithm with First-fit Partitioning

The problem of scheduling periodic tasks has been studied extensively since its first introduction by Liu and Layland. They proposed utilization bound, called L&L bound, to guaranty feasibility of task set under rate monotonic policy on a single processor. Recently, some papers have used semi-partitioned scheduling to increase overall system utilization. In semi-partitioned scheduling, most tasks are statically assigned to processors, while a fewtasks are split into several subtasks and each assigned to a different processor. In this paper a new semi-partitioned scheduling algorithm called FFDRM, which is based on delayed rate monotonic, is proposed for multiprocessor systems by which the system utilization ofmost of processors is raised up to L&L bound. Delayed rate monotonic is an improved version of rate monotonic. It is proved that the lower bound of feasibility of a system which includes two tasks under delayed rate monotonic is equal to one. FFDRM uses this ability of delayed rate monotonic to achieve higher system utilization than previous works. The feasibility of tasks which are partitioned by FFDRM is formally proved.

Keywords:

real-time embedded systems , semi-partitioned scheduling , first-fit allocation , delayed rate monotonic algorithm