Contrast Research And Application Of SQL And NoSQL

Nowadays with the rapid development in science and technology, the informatization degree of society become increasingly higher, and the traditional relational model database has been unable to meet the requirements of data processing for the information society. The development of new database technology is becoming more and more rapidly and urgent. As non-relational database(NoSQL) impacts constantly on traditional relational database, the voice of using NoSQL to take place of the relational database is progressively louder. Since the most comparisons and analyses of the two kinds of databases are concentrated in the category of the theory and concept of them, quantitative comparisons and analyses with concrete experiment data were rare. This paper researches and analyzes the performance differences between the two kinds of databases by correlation experiments, and it can provide reliable experimental basis for the choosing suitable database in practical application. Then it is researched that an establishment of a unified query abstraction layer for SQL and NoSQL database.In this thesis, it can be drawn the following conclusions by comprehensive comparison the relational database to the non-relational database, and doing experimental testing of their main representatives Microsoft SQL Server and Mongo DB to research and compare the performance of them.(1) As the growth of the data, Mongo DB is generally better than SQL Server on performance.(2) When inserting the data, the Mongo DB is better than SQL Server on performance, but if SQL Server uses SQL commands directly to replace the entity framework, its performance can be greatly improved.(3) When updating the data, Mongo DB is better than SQL Server if the query involves the primary key field; but if the query involves the non-key field, SQL Server performs better than Mongo DB.(4) When selecting the data, the performance of Mongo DB is significantly better than SQL Server if the query involves the key field or a join operation; but if the query involves the non-key field, the result would be the opposite. Moreover, in case of aggregate functions like calculating the average value, SQL Server performs better than Mongo DB with map-reduce method. However, when Mongo DB utilized its new aggregate framework, Aggregation Pipeline, it showed a significant improvement which resulted in even a slightly better performance than SQL Server. The outcome of this experimental research can objectively manifest that when people choosing the database in practical applications, the choice Mongo DB is feasible in terms of objective specific performance. At last, for scenarios that application data stored separately in relational database and non-relational database, an establishment of a unified query abstraction layer for SQL Server and Mongo DB database is proposed, which provides the theoretical basis for the realization of the unified query.