大数据系统综述

Abstract

With the development of the latest technologies, a large amount of data is generated from various domains (such as optical observation and control, healthcare, sensors, user-generated data, Internet and nancial companies, supply chain systems, etc.) during the last two decades. (A more appropriate description could be nite data, e.g., in the application of optical observation and control, data are continuously generated, cre-ating a data disaster.) The term of big data is coined to capture the profound meaning of this emerging trend.Compared with traditional data, big data exhibits some unique characteristics besides the sheer volume, such as commonly un-structured data and more real-time analysis requirements. The development of big data calls for new system architectures for data storage and large-scale data processing mechanisms. In this paper, we present a literature survey of big data analytics. Firstly, the de nition of big data and big data challenges are presented. Secondly, a systematic framework to decompose big data system into four sequential modules, namely data generation, data acquisition, data storage, and data analytics, which form the value chain for big data, is proposed. A detailed survey of numerous approaches and mechanisms related to each module, from research and industry communities is discussed. Finally, some evaluation benchmarks and potential scienti c problems in big data systems are outlined.