Informed Schema Design for Column Store-Based Database Services

Abstract

While database schema options in relational database management systems are few and have been studied for decades, little effort has so far been devoted to NoSQL column stores. Today, schema design for column stores is still based on the gut feeling of the application developer instead of being approached systematically. This is even more critical as "good" schemas in column stores do not only depend on the data model of the application but also on the queries on that data: Poor schema design will either lead to a situation where not all queries can be answered or where some queries will show really poor performance. In this paper, we propose a systematic and informed approach to database schema design in NoSQL column stores by means of automated schema generation and application-specific schema ranking.