层次化计算范式数据中心任务调度方法  

Data center task scheduling method oriented to hierarchical computing paradigm

作　　者：胡荣 马军生 HU Rong;MA Jun-sheng(Department of Intelligent Technology,Geely University,Chengdu 641402,China;Department of Information and Communication,National University of Defense Technology,Wuhan 430000,China)

机构地区：[1]吉利学院智能科技学院,四川成都641402 [2]国防科技大学信息通信学院,湖北武汉430000

出　　处：《计算机工程与设计》2022年第9期2511-2518,共8页Computer Engineering and Design

基　　金：吉利学院校级重点科研基金项目(2021KY0202)。

摘　　要：为解决异构层次化计算范式下,分布式应用的总延迟优化问题,提出一种数据中心任务调度方法。设计延迟感知的Power-of-D算法用于子任务的在线分发及调度,将经典的Power-of-D范式扩展到层次化计算范式中,利用表象处理速度进行问题建模描述,实现任务路由时延、处理时长和排队时长的联合优化。基于真实系统场景,对提出的方案进行了评估,评估结果肯定了方案的有效性。To solve the problem of total delay optimization of distributed applications under the heterogeneous hierarchical computing paradigm,a data center task scheduling method was proposed.The delay-aware Power-of-D algorithm was designed for the online distribution and scheduling of subtasks.The classic Power-of-D paradigm was extended to the hierarchical computing paradigm,and the representation processing speed was used to describe the problem modeling,to realize the joint optimization of task routing delay,processing time and queuing time.Based on the real system scenario,the proposed scheme was evaluated.The evaluation results confirm the effectiveness of the scheme.

关 键 词：层次化计算范式 分布式应用 延迟优化 调度方法 表象处理速度 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]