Benchmark Study For Non-relational Database And Achieve

Along with the information technology being applied to all aspects of human dailyproduction and life, a variety of modern applications such as Sensors, RFID, electronicmonitoring, financial applications, telecommunications applications, Internetapplications and all other applications generate huge unstructured data.Traditional andrelational databases have become inefficient to store and process enormous amount ofunstructured data. To solve this problem,NOSQL database is proposed by people fromresearch field and industry. After that, NOSQL database gets a rapid development andmeets with success in many applications. As long as the deep research and widespreaduse, NOSQL database plays more important roles in human life.The NOSQL database does not have a clear and unified standard.Each of them hastheir own features on system architecture, data model,data access interface,operationmode, etc. The best way to evaluate the adaptability of NOSQL database is benchmarktest, So this paper designs and implements an objective and impartial benchmark forNOSQL database.In this paper, studies on benchmark for NOSQL database are mainly focused onthe following aspects:First, the basic principles of NOSQL database are analyzed. The corefunctionalities and characters of NOSQL database are researched deeply. Thedevelopment trace, trend and current technologies of benchmark for NOSQL databaseare thoroughly analyzed. Based on the key-value data model, a benchmark is proposedfor testing NOSQL database.Secondly, after studies are developed from data type, data manipulation model, testindicators, metrics of indicators and basic tested model, a benchmark to NOSQLdatabase with key-value data model is proposed and implemented in this paper.Finally, based on the benchmark for NOSQL database which is proposed andimplemented in this paper, experiments give a comparison between two typical NOSQLdatabases that are HBase and Cassandra. The result of experiments showes thecharacteristics of these two databases and the effectiveness of work in this paper isproved.