The Multi-Operator Method: Integrating Algorithms for the Efficient and Parallel Evaluation of User-Defined Predicates into ORDBMS

There has been a long record of research for efficient join algorithms in RDBMS, but user-defined join predicates in ORDBMS are typically evaluated using a restriction after forming the complete Cartesian product. While there has been some research on join algorithms for non-traditional data (e.g. spatial joins), today's ORDBMS offer developers no general mechanism that allows to implement user-defined join predicates in an efficient way. We propose the multi-operator method to achieve this goal and show that it is suitable to implement joins with complex user-defined predicates much more efficiently than today. Our approach fits well into the architectural framework of current ORDBMS. A significant further benefit is that the multi-operator method, in our view, can serve as an enabling technique for the parallel execution of complex user-defined functions.