星载异构计算环境下的能耗优化任务调度

Real-Time Task Scheduling for Space-Oriented Computing on Heterogeneous System

针对航空航天探测等要求高可靠性、强实时性、低功耗的尖端领域,基于新型的异构多核计算平台对全迁移策略的支持,提出了能耗优化的实时任务调度算法。算法主要分为2部分:负载分配和任务调度。负载分配确定每个任务在不同的处理器集群上工作量的分配比例,并对系统的能耗进行优化。任务调度在负载分配的基础上,采用时间片划分的思想对已分配的任务进行合理调度,保证每个任务都能满足系统约束条件。将此算法与现有的SA算法和Hetero-Split算法进行对比,实验结果显示:此算法比Hetero-Split算法在寻求可行性任务调度方案方面相当,但强于SA算法;在能耗比方面,此算法比Hetero-Split算法在系统总能耗方面有大幅度的降低,降低比率约为23%~24%。

Aiming at the cutting-edge fields such as aerospace exploration which require high reliability,hard realtime performance,and low power consumption,a real-time task scheduling algorithm for energy optimization is proposed based on the two-type heterogeneous platform which supports full migration strategy. The algorithm is mainly divided into two parts,i.e.,load distribution and task scheduling. Load distribution,named EB-Split,determines the allocation of workload for each task on different processor clusters,and optimizes the energy consumption of the system. Task scheduling is based on the load distribution. It uses the concept of time-slicing to schedule the distributed tasks aptly,ensuring the system constraints. The EB-Split algorithm is compared with the existing simulated annealing(SA) algorithm and Hetero-Split algorithm through experiments. The results show that the EB-Split algorithm is equivalent with the Hetero-Split algorithm while stronger than the SA algorithm in seeking a feasible task scheduling scheme. For energy consumption,the EB-Split algorithm is much better than the Hetero-Split algorithm. Compared with the Hetero-Split algorithm,the EB-Split algorithm can reduce the total energy consumption of the system with a reduction ratio of 23%~24%.