Optimizing the MAX/MIN of SQL window functions

With the growing number of Internet users, the Internet application gradually enters the era of big data. How to store and analyze the big data becomes a problem in the Internet application. Window Function, as a solution of data analytics in the field of relational database, makes it take the place of Self Join and Sub Queries to complete the traditional complex queries with its subtle semantic characteristics and is widely used in the current enterprise data management and analysis in the Internet application. Under the background of big data, Window Function has the bottleneck in the face of such demand as high throughput and real-time response. In this paper, we detail the two-phase evaluation framework from the perspective of executor in the Relational Database Management System and design a new algorithm dependent on Temporary Window for the MAX/MIN Window functions contraposing the original framework in PostgreSQL. We design a query cost analysis model and theoretically proved the performance of the algorithm. We conduct the performance comparison between the new algorithm and the existing commercial database SQL Server and verify the effectiveness of the proposed algorithm.