基于可扩展计算平台的大图数据管理

Big Graph Management Based on Scalable Computing Platforms

随着数据采集和存储技术的发展,社交网络、生物信息科学、交通导航等领域中出现了规模庞大、内部结构复杂、查询需求多样的大图数据。传统基于单机内存的图处理方法无法满足大图数据管理需求。可扩展计算平台的发展为大图数据管理提供了可行的技术方案。本文首先分析了大图数据之上的不同类型查询,重点探讨了基于关系数据库、基于MapReduce计算框架、基于BSP(Bulk Synchronous Parallel)计算模型和基于第三方外包服务器的大图数据管理方法,并分析了未来可能的研究路线。

With the development of data collection and storage techniques, big graphs with massive size, complex intra-relationships and different graph queries, appear in social network, bio-informatics, and transportation network, etc. The classic graph algorithms which are suitable in memory of a single computer cannot be used to big graphs directly. The exploration of existing scalable platforms provides feasible solutions to manage big graphs. In this paper, we first investigate different kinds of graphqueries, and then analyze the advantages and disadvantages of different approaches to management of big graphs using various platforms, including relational database, MapReduce framework, BSP （Bulk Synchronous Parallel） model and third party outsourced servers, and finally outline the future work.