Optimizing Schedule Length for DAG Type Applications with Energy Consumption Constraint in Heterogeneous Computing Systems

Yu He,Jinchao Chen,Chenglie Du,Pengcheng Han

DOI: https://doi.org/10.1109/pic50277.2020.9350786

2020-01-01

Abstract:Schedule length minimization of a parallel application with precedence constrained tasks in heterogenous computing systems is always a research hotspot of distributed computing. Unfortunately, accelerating the execution of tasks requires faster processors and consequently produces more energy. Users are faced to a time-energy tradeoff in selecting processors and execution frequencies for tasks. A common way to address this trade-off is to minimize schedule length under energy constraints. Up to now only a few works have been done to address this problem. Nevertheless, the energy preassignment methods of these works are unfair for low priority tasks in various degrees. Furthermore, existing works pay little attention to the priority assignment strategy, which can be critical for task scheduling. Hence, we proposed a novel approach named DAG Scheduling with Energy Consumption Constraint (DSECC) to solve this problem. DSECC combines several priority assignment strategies to assign priorities for tasks, and preassigns energy for tasks based on the average energy consumption of tasks on different processors. A large number of experiments based on randomly generated DAG show that compared against state-of-the-art algorithms, our proposed algorithm can obtain smaller schedule length and generate less energy consumption.