In-Memory Database Optimization Using Statistical Estimation

Abstract

Existing dictionary compression algorithms are notable to preserve the property of direct access, thereby leading to a severe hit in performance. In this paper, we propose a prudent approach to format the existing dictionary tables to achieve significantly better performance levels, without losing direct access. The approach is based on reducing the unused space in a dictionary column and works on simple premises derived from statistics. This paper further discusses a technique to improve performance of the attribute vector so created, after employing the proposed dictionary approach.