Development and Implementation of Complex Relational Data Processing Operators in the PostgreSQL DBMS Dynamic Query Compiler

As performance of main and external memory grows, alongside with decreasing hardware costs per GB, program execution time is more determined by CPU characteristics and the way it is utilized. The comparison of traditional disk-oriented database management systems initially developed in the previous century, and modern in-memory solutions that are highly optimized for CPU as well as a considerable volume of RAM, shows a significant gap in data processing performance. In this work, we examine the method of dynamic query compilation as the way of query execution speed improvement in disk-oriented DBMS PostgreSQL. To improve the quality of generated machine code, we utilize a specific transformation method to convert existing pull-based relational operators into push-based model. The transformation is carried out by decomposition of existing operators’ algorithms into special interface functions. The application of the decomposition method we demonstrate on several data processing operators of different complexity.