Data management for virtual reality visualization of large flow datasets

Traditional flow visualization systems are limited by the lack of appropriate data management support, and require that the entire data to reside in main memory. The goal of this research is elaboration of a high-performance software system that manages the large quantities of data stored on a prevalent RAID class storage system and moved to main memory in time to support virtual reality visualization.; The software system makes use of an octree-based model to create new datasets at various levels of detail, which retain as many essential features of the original data as possible but are small enough to be loaded one chunk at a time into main memory. The new octree datasets are organized as a skeleton for systemic access, which supports multiresolution visualization so that the active region of interest is always displayed at a higher resolution than other regions.; Octrees are efficiently partitioned, organized and stored on disk. An index tree for disk-resident octree data is provided to accelerate the search process at run-time. This separate structure provides essential property information but is lightweight so that it can be traversed quickly by the visualization program, which retrieves the set of data that is contained in all index entries needed for a given time step. Furthermore, prefetching is used to mitigate disk access time. The system minimizes the number of disk operations, data movement, and network communication requirements by storing only that part of the flow field that makes up the display objects which, in general, could be orders of magnitudes smaller than the entire flow field data. Consequently, octree methods improve the overall system performance, yielding frame rates suitable for interactive visualization on common desktop graphics workstations.; The UCLA's Virtual Aneurysm and Virtual World Data Server projects serve as the front-end and back-end application testbeds, respectively. The visualization task we tested is the use of a cutting plane to investigate the speed of blood flow within the aneurysm in any orientation. Practical experiments indicate that the techniques developed in this system provide the basic building block for efficient implementation of large complex visualization systems.