Further analysis on blocking time bounds for partitioned fixed priority multiprocessor scheduling

Abstract

Partitioned fixed priority scheduling is one of the most comprehensively chosen predictable scheduling in practice. The FMLP+ developed in recent years is a better protocol, which ensures asymptotically optimal O(n) maximum priority-inversion blocking. The constraints under several protocols besides the FMLP+ can be exploited to gain the bound on maximum blocking time. However, the blocking time bounds may be pessimistic under the FMLP+ on the ground that shared resources local to a processor do not incur priority-inversion blocking in some cases. It is possible that a schedulable task set is judged as unschedulable because of the pessimistic values. Based on our analysis, a few constraints was added to compute the maximum blocking time of each task, and then its worst-case response time. The results of our experiments show less pessimism than the existing ones. Meanwhile, we also demonstrate the usefulness of the conclusion that global resource sharing should be transformed into local one where possible.