Toward RDB to NoSQL

Abstract

The emergence of applications dealing with structured, semi structured, and non-structured data created demands on new data storage systems. The relational model, widely used to store data from diverse applications, does not meet all the imposed scenarios. In response, NoSQL databases have emerged as an option. As a consequence, new approaches for converting the relational model to NoSQL models have been created. However, most of them target a specific NoSQL model and provide little (or none) support for customization for transformations with different cardinalities. In this paper we present a novel approach to convert relational databases (RDB) to document and column family NoSQLs. Our approach receives as input a set of directed acyclic graphs (DAG) representing the target NoSQL model. The DAGs are used to generate commands to transform the input RDB data. The approach supports different cardinalities, and the commands can be customized. We have developed a tool to interpret the DAGs and that supports different transformation strategies. We performed experiments to validate our solution, with different configurations, where conversions were carried out to document and column-family storage.