Research On Open Scalable Relational Data Model And Data Partition Strategy

With the Internet becoming an indispensable part of the society, the generatingspeed of data from all walks of life has reached an unprecedented level. The conceptBig Data is brought up upon this background and attracts increasing attention. Infront of the challenges, big data technologies arise at the history moment. To someextent, big data technology is aimed at fetching valuable information from varioustypes of data. Among these technologies, database and related research on datamodels and data storage patterns address the issues of how to describe, organize andstore large data sets efficiently, providing qualified access capacity for applications.This kind of research strongly supports the upper levels of database construction, dataaccess and mining.This paper describes the background, significance of big data research work andtypical features of big data, analyzes current research status of data model and datapartition strategies. Then by considering some shortcomings of traditional databasessuch as limited support for data types, low concurrent query performance and lack ofhorizontal scalability, this paper puts forward an Open Scalable Relational DataModel (OSRDM), which provides an open support for varieties of data types andenables fully horizontal scalability on basis of keeping and extending relationaldescriptive power of traditional relational data model, catering to the characteristicsof big data such as volume, variety and velocity. On top of this, aimed at constructinga storage system with high scalability, high reliability and easy deployment acrossdata centers, this paper analyzes the existing data partition and redundant mechanismsand proposes a Consistent Hash based Adaptive Flexible Redundant Data PartitionStrategy (CHAFRAS). This strategy combines Consistent Hash Algorithm withreplication-based redundant mechanism, maximizes the high horizontal scalability ofkey-value-based data models, guarantees the data reliability and improves concurrentaccess performance. Evaluation is divided into two parts: one is the analysis of functional features,and the other is the quantitative experiment. In each part, current popular models andstrategies are introduced to compare with the proposed models and strategies. Theresults show the superiority of OSRDM and CHAFRAS.