Data Mining Implementations for High Performance Computing to Support MDA

"Collaborative applications involve the acquisition, processing, and delivery of very large volumes of remote sensing and related data for a variety of applications. In order to handle this data, collaborative environments must manage, store, search for andretrieve data across highly distributed and networked topologies. This effort will provide a technology that supports distributed data management by developing software libraries that consists of data access interfaces and search algorithms that facilitateparallel data analysis and mining. These libraries will be integrated with existing simulation technologies such as SPEEDES and RAM Laboratories WarpIV simulation engine to provide a parallel and distributed data mining capability. This effort will takeadvantage of existing persistence and save/restore mechanisms to support basic storage and retrieval of data. More complex search mechanisms that traverse a parallel hierarchical grid data structure will then be employed to coordinate searches usingmassively parallel high-performance interest management. The resultant technology can be applied to both government-based and commercial collaborative applications. This SBIR is viewed as having excellent commercial potential. Naturally, this SBIR and thedistributed data mining technology it will develop have far reaching implications for BMDO. The data access interface and algorithm libraries will allow for better performance and data