Models and Adaptive Architecture for Smart Data Management

Abstract

Organizations, companies and Web platforms hold large amounts of unused data. These data are trapped in separate data sources, locked up in legacy formats and only reachable through several different protocols, making usage difficult. It is therefore necessary to manage this multiplicity of data sources in order to build a solution able to combine this multi-origin data into a coherent smart data set. We define a meta-model and models to describe data source diversity in a flexible way. We therefore propose an adaptive architecture that generates data integration workflows at runtime. We evaluate our approach to offer scalability, responsiveness, and dynamic and transparent data source management. We apply our approach in a live scenario from a French company to show how it adapts to industrial needs and facilitates smart data production and reuse. This paper describes our models and strategies and presents our resource-oriented architecture.