CCA并行构件程序分布式负载均衡方法  

Distributed load balancing method for CCA parallel component applications

作　　者：彭云峰 刘家磊 石聪明 高国伟 Peng Yunfeng;Liu Jialei;Shi Congming;Gao Guowei(School of Software Engineering,Anyang Normal University,Anyang Henan 455000,China;School of Computer Engineering,Hubei University of Arts&Science,Xiangyang Hubei 441053,China;Hubei Key Laboratory of Power System Design&Test for Electrical Vehicle,Hubei University of Arts&Science,Xiangyang Hubei 441053,China)

机构地区：[1]安阳师范学院软件学院,河南安阳455000 [2]湖北文理学院计算机工程学院,湖北襄阳441053 [3]湖北文理学院纯电动汽车动力系统设计与测试湖北省重点实验室,湖北襄阳441053

出　　处：《计算机应用研究》2024年第12期3793-3800,共8页Application Research of Computers

基　　金：河南省重点研发与推广专项资助项目(科技攻关)(242102210179,232102210026,232102210036);高原与盆地暴雨旱涝灾害四川省重点实验室开放研究基金资助项目(SZKT202104);安阳市科技计划资助项目(科技攻关)(2021C01GX015,2022C01NY020);湖北省自然科学基金计划资助项目(一般面上项目)(2023AFB585);湖北省科技重大专项资助项目(2023BCA006);湖北省优势特色智能与新能源汽车学科群。

摘　　要：现有的并行构件程序大多采用静态或者集中式的动态负载均衡策略,性能优化效果并不理想。通过分析CCA(Common Component Architecture)并行构件程序的运行机制,提出一种动态的、分布式的并行构件程序负载均衡方法。使用基于面向对象机制的方法建立计算节点的类库,通过数据流分析管理不同并行构件调用之间的依赖关系,将不存在依赖的并行构件实例作为可以并行执行的任务分配给计算平台上的不同计算节点执行。实验表明,相对于已有的静态或集中式的负载均衡策略,能更好地利用计算资源,同时避免可能由管理节点造成的瓶颈效应,有较小的负载均衡开销,能取得更好的性能效果,具有较好的可扩展性。Existing parallel component applications use static or centralized dynamic load balancing strategy,and the perfor-mance optimization effect is not satisfactory.This paper analyzed the running mechanism of CCA(Common Component Architecture)parallel component applications,and proposed a dynamic and distributed load balancing method.The method established a class library for computing nodes using an object-oriented mechanism based approach.It managed the dependency relationships between different parallel component calls through data flow analysis,and assigned parallel component instances that do not have dependencies to different computing nodes on the computing platform as tasks that can be executed in parallel.Experiments show that,compared to existing static or centralized load balancing strategies,it can better utilize computing resources while avoiding bottleneck effects that may be caused by management nodes,with smaller load balancing costs,can achieve better performance,and has good scalability.

关 键 词：并行构件 性能优化 负载均衡 并行计算 资源管理 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]