Decrease Iteration Time Deterministic Cyclic Scheduling for Real-time Periodic Tasks

Abstract

Considering the characteristics of optimization theory and periodic tasks, this paper proposes a novel algorithm named as Decrease Iteration Time Deterministic Cyclic Scheduling (DITDCS). This is a soft real-time task allocation scheme based on optimization theory in a heterogeneous multi-core environment and uses task periodicity and DVFS (Dynamic Voltage and Frequency Scaling) to optimize the scheduling algorithm for computing time and system energy consumption. This algorithm creates a buffer to save a certain number of periodic tasks arriving and obtains the task allocation matrix in the buffer through the task allocation algorithm based on optimization theory. At the same time, after each task is assigned, the number of iterations in the calculation is reduced by updating the allocation matrix to improve the performance of the algorithm. Due to the characteristics of periodic tasks, the scheduling cycle with a certain length can be regarded as a scheduling template, which is used in subsequent scheduling to further reduce the computational overhead, and the effect is significant in a large-scale task set. Through experimental analysis, the proposed algorithm can achieve better energy consumption under the guarantee of task completion time and limited delay, and the overhead is also lower than the same type of algorithm.