On-Demand Service-Based Big Data Integration: Optimized for Research Collaboration

摘要

Biomedical research requires distributed access, analysis, and sharing of data from various disperse sources in the Internet scale. Due to the volume and variety of big data, materialized data integration is often infeasible or too expensive including the costs of bandwidth, storage, maintenance, and management. Obidos (On-demand Big Data Integration, Distribution, and Orchestration System) provides a novel on-demand integration approach for heterogeneous distributed data. Instead of integrating data from the data sources to build a complete data warehouse as the initial step, Obidos employs a hybrid approach of virtual and materialized data integrations. By allocating unique identifiers as pointers to virtually integrated data sets, Obidos supports efficient data sharing among data consumers. We design Obidos as a generic service-based data integration system, and implement and evaluate a prototype for multimodal medical data.