Toward building RDB to HBase conversion rules

Abstract

Cloud data stores that can handle very large amounts of data, such as Apache HBase, have accelerated the use of non-relational databases (coined as NoSQL databases) as a way of addressing RDB database limitations with regards to scalability and performance of existing systems. Converting existing systems and their databases to this technology is not trivial, given that RDB practitioners tend to use the relational model design mindset when converting existing databases. This can result in inefficient NoSQL design, leading to suboptimal query speed or inefficient database schema. This paper reports on a two-phase experiment: (1) a conversion from RDB to HBase database, a NoSQL type of database, without the use of conversion rules and based on a heuristic approach and (2) an experiment with different schema designs to uncover and extract conversion rules that could be useful conversion guidelines for the industry.