Flexible Relational Data Model – A Common Ground for Schema-Flexible Database Systems

Abstract

An increasing number of application fields represent dynamic and open discourses characterized by high mutability, variety, and pluralism in data. Data in dynamic and open discourses typically exhibits an irregular schema. Such data cannot be directly represented in the traditional relational data model. Mapping strategies allow representation but increase development and maintenance costs. Likewise, NoSQL systems offer the required schema flexibility but introduce new costs by not being directly compatible with relational systems that still dominate enterprise information systems. With the Flexible Relational Data Model (FRDM) we propose a third way. It allows the direct representation of data with irregular schemas. It combines tuple-oriented data representation with relation-oriented data processing. So that, FRDM is still relational, in contrast to other flexible data models currently in vogue. It can directly represent relational data and builds on the powerful, well-known, and proven set of relational operations for data retrieval and manipulation. In addition to FRDM, we present the flexible constraint framework FRDM-C. It explicitly allows restricting the flexibility of FRDM when and where needed. All this makes FRDM backward compatible to traditional relational applications and simplifies the interoperability with existing pure relational databases.