Description Marius Slavescu 2007-01-02 12:26:38 EST

This feature should provide a service oriented data management infrastructure.

This feature should also be the answer to scalability related requests.

The infrastructure could be based on a two layer design where the top layer provides the data domain abstractions and the lower layer represents the data store abstractions. The application code is supposed to interact only with the top layer where the top layer will transparently interact with the lower layer.

The main capabilities of this infrastructure should be:

- allow high performance read/write and analysis of large data sets (limited by the disk availability).
- provide a simple data representation as the default implementation (should cover the current analysis, viewing, and reporting operations for most data domains, initially it may not support completely the complex filtering found in some existing scenarios)
- provide support for extensibility for both data domain level and data store level interactions
- allow any concrete data store to participate in the data persistence layer (should not be limited to RDB concrete implementations)
- should also work on a single machine with embedded data stores (also allow disconnected mode)

This is intended to replace the current EMF based implementation, basically to achieve a simple and more scalable view of the data operated by the TPTP UI, analysis and reporting components.

The event processing (data normalization) will be covered in the component that covers load data operations (basically the event loaders will interact with this component).

Gradually we will move TPTP functions to use this infrastructure (as this will become concrete), backward compatibility should be maintained.