Transforming a multi-value database system into a relational database system for faster querying

Abstract

When the data is fluid and relationships among attributes are getting complex and are changing rapidly over time, a Multi-value Database Management System (MVDBMS) can provide a great deal of support to handle that. MVDBMSs can handle attributes taking a list of values unlike a Relational Database Management System (RDBMS) that can only accept single-valued attributes to conform first normal form (1NF). It is suitable to use multi-valued database in case where a list of values can be assigned to a single attribute in a particular record. In this case, MVDBMS can save a lot of storage space and it is quite efficient to retrieve a specific record. However, when it comes into retrieving data based on Select-Project-Join (SPJ) queries, a multi-valued database cannot handle these queries in a straightforward manner. In this paper, we propose techniques to transform a MVDBMS into a RDBMS for faster querying and retrieval. During the transformation, we carefully considered the trade-off in between execution time of queries and storage space consumed by data. Therefore, we introduce two novel techniques: one of them will generate database schema conforming to 3NF and the other will be in 1NF. We also evaluate the performance of these two techniques for different types of queries.