Unified Views for Querying Heterogeneous Multi-model Polystores

Data storage, in various SQL and NoSQL systems brings complexity to data querying when entities are fragmented because data is not always stored in the same system, plus heterogeneous structures can appear for entities. A unique query language is not sufficient to address data distribution and heterogeneity. Considering vertically distributed data, this work implements a framework capable of rewriting a user query addressed over a unified view to access all data and provide results with transparency. Our framework works with a conceptual model producing unified views to guarantee polystore querying without having to know data distribution nor data heterogeneity. It complements the initial query with intermediate operations. It is applied on an e-commerce scenario (UniBench benchmark) distributed vertically between relational and document-oriented databases. Performance results and the low impact of query rewriting process are illustrated in this work.