一种多 DAG 任务共享异构资源调度的费用优化方法

摘要

Aiming to the new problems of cost optimization for concurrent multiple DAG (Directed Acyclic Graph) workflow sharing heterogeneous resource on the basis of maximizing overall throughput of DAGs scheduling ,we propose an algorithm called PDTC (based on the Probe of the Total Cost Decrease )to minimize the total cost of these DAG while meeting a user-defined dead-line .Last experiments demonstrate that these algorithms and methods can improve the related performances .%针对多DAG（Directed Acyclic Graph）共享一组异构资源在调度吞吐量最大化基础上的费用优化问题，本文提出了一种基于总费用变化量探测的费用优化算法PDTC （based on the Probe of the Total Cost Decrease ），目的在于尽可能降低有优化条件的多个DAG的总费用。实验表明，该算法不仅能使得各DAG充分利用期限内的冗余时间，也能够在一定程度上降低多个DAG调度执行的总费用。