Integration of an In-Memory Column-Oriented Database Engine as Back-End of a BI Application

Sammanfattning: The benefits of Business Intelligence (BI) activities are widely recognized and arouse more and more interest nowadays. In particular, self-service analytics is getting more and more popular. Enabling end users to process their data by themselves allow them to be more reactive to business changes by making better decisions faster. One of the challenges in this area is to provide tools easy to use with a high computational performance to enable end users who often aren’t developers to process their data by themselves in a simple and fast manner. The BI application that we will call BITX belongs to that category. This thesis presents the integration of an in-memory columnoriented database engine that we will call DBEX as backend for this application. The purpose of this experiment is to optimize BITX’s performance by providing it with a lightweight back-end well adapted to BI analysis. The architecture and implementation of the driver developed for DBEX are detailed. The query execution mechanism and SQL functions representation in the engine are studied and performance optimizations for SQL functions implementation are proposed. The system performance is then analyzed following two axis: DBEX’s memory consumption and the respone time of the subsystem {DBEX+driver}. Improvements made further to the performance evaluation and speeding up some of the driver’s operations are presented.