大数据管理的偶图模型

Bigraphical Model for Big Data Management

大数据管理是大数据研究的一个重要方面.随着数据量的增大,大数据管理的复杂性成指数级增加,大数据生命周期思想是解决这一复杂性的有效方法之一.目前大数据管理大都基于非形式化和半形式化方法,缺少严密的形式化方法.偶图不仅是形式化建模工具,而且其图形化表达非常直观.本文采用偶图以及偶图反应系统相关理论,并结合大数据生命周期思想,建立大数据管理的偶图模型.然后,用偶图模型重写开源分布式文件系统Fast DFS和MapReduce编程模型,并用带演算的MapReduce偶图模型改进交互数据处理效率低的弱点.两个案例表明提出的大数据管理形式化方法的有效性.

Big data management is an important aspect of big data research. As the volume of data grows, the complexity of big data management is increasing in an exponential way. Big data life cycle method is one of the effective approaches to solving this complexi- ty. At present,big data managements are mostly based on non-formal and semi-formal methods, lacking of rigorous formal methods. Bigraph is not only a formal modeling tool, but also has intuitive expression capability. Combined the big data life cycle with theories for the bigraph and bigraphical reaction system, we establish a bigraphical model to manage big data in this paper. Then, we try to ex- press open source distributed file system FastDFS in formal bigraphical model and use bigraphical model to rewrite programming mod- el MapReduce. The model shows that bigraphical model with calculation of big data management can overcome the weakness of Ma- pReduce during processing interactive data. The two cases presented in this paper show that the proposed formal method for big data managements is effective.