Mapping geographic conceptual schemas to nosql graph databases

Abstract

Geographic conceptual modeling, as traditional conceptual modeling, is a basic activity in the design of geographic applications. Geographic conceptual models provide primitives to represent the geometry and the topology of geographic data, which are generally stored in Geography Markup Language (GML) documents or in spatial databases. Database design at the conceptual level must be independent from the underlying mechanisms that implement the database. A good conceptual schema can be mapped to various physical schemas, so that the application can benefit from the best characteristics of the database management system used in its implementation. It also allows total or partial schema reuse, since the modeled part of the real world can be perceived in a similar manner by different applications. However, conceptual models that consider geographic data and applications, such as the Object Modeling Technique for Geographic Applications (OMT-G), include a wider variety of representation primitives for classes and relationships. At the same time, a number of alternative database management systems, which implement novel representation paradigms and data structures, have recently arisen with the generic name of NoSQL (or not only SQL). This dissertation defines rules and algorithms to implement the automatic mapping between OMT-G schemas and NoSQL graph-based logical and physical schemas. A comprehensive example of the proposed process is presented. Conclusions indicate that mapping geographic conceptual schemas to hybrid (relational and graph) physical schemas may be desirable in the future, in order to combine the best characteristics from each implementation alternative.