The   concept is demonstrated by the overall architecture shown below.


Data sources are multiple streaming data sources, as well as archival data sources.

In-situ processing components aim to compress and integrate where appropriate data-in-motion from streaming sources in communication efficient ways computing single and multi-streaming data synopses at high rates of data compression. 

Data transformation components aim to covert data in (a) single and multi-streaming data synopses, (b) archival data and (c) results from the datAcron higher levels analytics components to a common form.

Data integration component interlinks semantically annotated data using link discovery techniques for automatically integrating and interlinking data from disparate sources. Integrated data are provided to the analytics components in real-time, while they are also stored in parallel stores.

Spatiotemporal query-answering component provides parallel query processing techniques for spatiotemporal query languages. Interlinked data are stored in RDF stores, using sophisticated RDF partitioning algorithms in domain specific, spatial and temporal dimensions.

Data analytics components include trajectory and complex event recognition and forecasting, as well as visual analytics. These consume the data provided by the data integration component: Synopses computed by the bottom layer, being integrated with archival data.