A Novel Dynamic Allocation and Scheduling Scheme with CPNA and FCF Algorithms in Distributed Real-time Systems

Abstract

This paper presents a new scheme for jointly and dynamically allocating and scheduling tasks with both periodic and aperiodic time constraints in distributed real-time system. The scheme includes a new allocation algorithm, a new scheduling algorithm and a heuristic scheduling algorithm. The new allocation algorithm is called classified-processing-nodes allocation (CPNA). In CPNA algorithm, we divide the processing nodes (PNs) into two sets: set A and set P. Set A is used to accept aperiodic real-time tasks. Periodic real-time tasks are dispatched to set P. The partition to PNs is not fixed but flexible and auto-adaptive. The new dynamic scheduling algorithm is called first committed first (FCF) algorithm, which is used to schedule periodic real-time tasks. In our scheme, we use a heuristic scheduling algorithm to schedule aperiodic real-time tasks. To evaluate our scheme, we compare CPNA with equal-processing-nodes allocation (EPNA), FCF with RMS through experiments. The experimental results show that our scheme with CPNA and FCF is more efficient than the scheme with EPNA and RMS.