This research focuses on a comprehensive approach to achieving individual and collaborative interactions for extremely complex simulation data. Such datasets would overwhelm existing systems. They often share many common attributes in their organization, data evolution and dependencies, and patterns of shared, coherent access by human or automated agents. These commonalties can be captured through a suitable set of abstractions, and exploited to achieve fundamental improvements in methods for manipulating, sharing, and understanding such broad collections of data. Specifically, the project is aimed at constructing a system for interactive visualization of extremely complex datasets, by developing techniques for: (1) query factoring, (2) predictive fetching, (3) detail degradation, (4) conservative algorithms, (5) hybrid cost metrics, and (6) an interactive tool infrastructure. The educational component addresses the potential for collaborative interactive techniques to improve pedagogy at the undergraduate, graduate, and professional levels, and performance evaluation at the undergraduate level.