基于Ray框架的分布式计算研究  

Research on Distributed Computing Based on Ray Framework

作　　者：侯伟 HOU Wei(School of Information and Control Engineering,Liaoning Petrochemical University,Fushun 113001,China)

机构地区：[1]辽宁石油化工大学信息与控制工程学院,辽宁抚顺113001

出　　处：《现代信息科技》2023年第23期65-68,共4页Modern Information Technology

摘　　要：文章旨在通过分布式框架Ray来计算数据量比较大的任务,使得任务能够更加高效地完成。指明了Ray通过两种核心概念实现分布式计算:Actor和Task。通过调用Ray库中的Remote方法并结合init方法指明CPU数量实现分布式计算。实验分为在相同数据量下不同CPU数量处理数据的快慢以及在不同数据量下不同CPU数量处理数据的能力。实验结果表明,数据量越大、CPU数量越多的分布式计算能力越强。因此,分布式计算适用于大数据下的并行计算。This paper aims to use the distributed framework Ray to calculate tasks with large amounts of data,enabling tasks to be completed more efficiently.It indicates that Ray implements distributed computing through two core concepts:Actor and Task.By calling the Remote method in the Ray library and combining it with the init method to indicate the number of CPUs,distributed computing is achieved.The experiment is divided into the speed of processing data with different CPU number under the same data volume,and the ability of processing data with different CPU number under different data volumes.The experimental results indicate that the larger the data volume and the more CPUs,the stronger the distributed computing power.Therefore,distributed computing is suitable for parallel computing under big data.

关 键 词：分布式计算 Ray框架 ACTOR TASK 大数据 CPU 

分 类 号：TP3-0[自动化与计算机技术—计算机科学与技术]