Optimizing scalar user-defined functions in in-memory column-store database systems

Abstract

User-defined functions such as currency conversion and factory calendar are important ingredients in many business applications. Since currency conversion and factory calendar are expensive userdefined functions, optimizing these functions is essential to high performance business applications. We optimize scalar user-defined functions by caching function call results. In this paper we investigate which method for function result caching is best in the context of in-memory column-store database systems. Experiments show that our method, which implements a function result cache as an array, combined with SAP HANA in-memory column store provides the high performance required by real-time global business applications.