Parallel I/O for 3D Volume Visualization

本项目采用并行I/O技术来减少体视化中的I/O延迟。 从诸如计算机断层扫描和3D超声的应用中产生的体积图像太大而不能容纳在随机访问帧缓冲器内。 因此，磁盘I/O是体绘制的关键路径。 该项目正在探索使用一种新的并行体绘制技术，该技术以消除并行磁盘阵列中I/O冲突的顺序遍历3D数据集。 因此，磁盘阵列的整个带宽可用于渲染应用程序。 在这个项目中，一个高性能的体积可视化服务器，使用这种想法正在设计，实现与现成的组件，并进行评估。 补充支持本科生的研究经验将允许高年级本科生通过实施和测量I/O密集型应用程序为该项目做出贡献。 理由

This project is applying parallel I/O technology to reduce I/O delay in volume visualization. Volume images that arise from such applications as computer tomography and 3D ultrasound are too large to fit within random access frame buffers. As a result, disk I/O is in the critical path for volume rendering. This project is exploring the use of a new parallel volume rendering technique that traverses 3D data sets in an order that eliminates I/O conflicts in a parallel disk array. Consequently, the entire bandwidth of the disk array is available to the rendering application. In this project, a high performance volume visualization server using this idea is being designed, implemented with off-the-shelf components, and evaluated. A supplement supporting Research Experiences for Undergraduates will permit advanced undergraduate students to contribute to the project by implementing and measuring I/O intensive applications. JUSTIFICATION