面向大数据分析的分布式矩阵计算系统研究进展

在大数据治理应用中,数据分析是必不可少的一环,且具有耗时长、计算资源需求大的特点,因此,优化其执行效率至关重要.早期由于数据规模不大,数据分析师可以利用传统的矩阵计算工具执行分析算法,然而随着数据量的爆炸式增长,诸如MATLAB等传统工具已无法满足应用需求的执行效率,进而涌现出了一批面向大数据分析的分布式矩阵计算系统.从技术、系统等角度综述了分布式矩阵计算系统的研究进展.首先,从发展成熟的数据管理领域的视角出发,剖析分布式矩阵计算系统在编程接口、编译优化、执行引擎、数据存储这4个层面面临的挑战;其次,分别就这4个层面展开,探讨、总结相关技术;最后,总体分析了典型的分布式矩阵计算系统,并展望了未来研究的发展方向.

Research Progress on Distributed Matrix Computation Systems for Big Data Analysis

As an essential part of big data governance applications,data analysis is characterized by time-consuming and large hardware requirements,making it essential to optimize its execution efficiency.Earlier,data analysts could execute analysis algorithms using traditional matrix computation tools.However,with the explosive growth of data volume,the traditional tools can no longer meet the performance requirements of applications.Hence,distributed matrix computation systems for big data analysis have emerged.In this paper,we review the progress of distributed matrix computation systems from technical and system perspectives.First,this paper analyzes the challenges faced by distributed matrix computation systems in four dimensions:programming interface,compilation optimization,execution engine,and data storage,from the perspective of the mature data management field.Second,this paper discusses and summarises the technologies in each of these four dimensions.Finally,the paper investigates the future research and development directions of distributed matrix computation systems.