ISSN 1000-1239 CN 11-1777/TP

• 系统结构 •

### 异构系统双关键级分布式功能的动态调度

1. 1(湖南大学信息科学与工程学院 长沙 410082);2(嵌入式与网络计算湖南省重点实验室(湖南大学) 长沙 410082);3(湖南省发展和改革委员会 长沙 410004) (llj1984109@qq.com)
• 出版日期: 2016-06-01
• 基金资助:
国家自然科学基金项目(61173036,61202102,61300039,61300037,61402170);国家“八六三”高技术研究发展计划基金项目(2012AA01A301-01);中国博士后科学基金项目(2016M592422)

### Dynamic Scheduling of Dual-Criticality Distributed Functionalities on Heterogeneous Systems

Liu Liangjiao1,2, Xie Guoqi1,2, Li Renfa1,2, Yang Liu3, Liu Yan1,2

1. 1(College of Computer Science and Electronic Engineering, Hunan University, Changsha 410082);2(Key Laboratory for Embedded and Network Computing of Hunan Province (Hunan University), Changsha 410082);3(Development and Reform Commission of Hunan Province, Changsha 410004)
• Online: 2016-06-01

Abstract: Heterogeneous distributed systems are mixed-criticality systems consisting of multiple functionalities with different criticality levels. A distributed functionality contains multiple precedence-constrained tasks. Mixed-criticality scheduling of heterogeneous distributed systems faces severe conflicts between performance and time constraints. Improving the overall performance of systems while still meeting the deadlines of higher-criticality functionalities, and making a reasonable tradeoff between performance and timing are major optimization problems. The F_DDHEFT(fairness of dynamic dual-criticality heterogeneous earliest finish time) algorithm is to improve the performance of systems. The C_DDHEFT (criticality of dynamic dual-criticality heterogeneous earliest finish time) algorithm is to meet the deadlines of higher-criticality functionalities. The D_DDHEFT (deadline-span of dynamic dual-criticality heterogeneous earliest finish time) algorithm is to allow the lower-criticality functionalities to be processed positively for better overall performance while still meeting the deadlines of higher-criticality functionalities, such that a reasonable tradeoff between performance and timing is made. Both example and extensive experimental evaluation demonstrate significant improvement of the D_DDHEFT algorithm.