CISE Research Instrumentation: Application and Evaluation of a Low Cost Parallel Processing Platform

9617384 Apon，Amy W. 皮肯斯三世，大卫R. 范德比尔特大学 CISE研究仪器：低成本并行处理平台 一笔仪器赠款将资助一个由高速互连连接的低成本计算机网络的组装。我们的低成本、高速网络将作为三个项目的试验平台，同时推进应用领域的研究，并展示网络作为并行处理平台的成本效益。 其中一个项目将侧重于功能性磁共振成像（fMRI），它表明大脑中积极处理信息的区域。目前的处理技术在处理大型数据集时可能需要数小时。 软件将在并行系统上重新实现，目的是减少处理时间，使功能磁共振成像更实用于研究大脑功能。 第二个项目将侧重于数据挖掘，即发现数据中“有意义的”模式。数据正在以远远超过我们分析能力的速度积累，网络将允许利用并行性来分析科学和工业数据。 第三个项目将集中在技术，如线程和数据放置和通信策略，提高性能的并行应用程序的高速网络。 实验平台将允许开发和评估各种技术。 分析模型将被用来概括跨可比架构的结果。

9617384 Apon, Amy W. Pickens III, David R. Vanderbilt University CISE Research Instrumentation: Low Cost Parallel Processing Platform An instrumentation grant will fund an assembly of a network of low-cost computers linked by high-speed interconnections. Our low-cost, high-speed network will serve as a test-bed for three projects, which will simultaneously advance research in the application areas and demonstrate the cost-effectiveness of the network as a parallel processing platform. One project will focus on functional magnetic resonance imaging (fMRI), which indicates areas of the brain that are actively processing information. Current processing techniques can take hours on large data sets. Software will be reimplemented on the parallel system with the goal of reducing processing time so that fMRI is more practical for research into brain function. A second project will focus on data mining, which discovers `meaningful' patterns in data. Data is being amassed at a rate that far exceeds our ability to analyze it. The network will allow parallelism to be exploited for analyzing scientific and industrial data. The third project will focus on techniques such as thread and data placement and communication strategies that improve the performance of parallel applications on the high-speed network. The experimental platform will allow various techniques to be developed and evaluated. Analytic models will be used to generalize the results across comparable architectures.