MRI: Enabling Research on Terabyte-Scale Datasets

该项目收购了一个集群，用于研究信息检索，网络流量分析，入侵检测和生物医学图像处理领域的TB级问题，通过为用户提供长时间的TB级存储访问，实现长期研究。在这个共享设施中，调查人员将合作利用一个包含20 TB磁盘存储的集群。此外，通过光纤通道连接的高性能TB磁盘子系统将作为大型磁盘存储的高速缓存。为了完成内存存储，数据迁移将利用现有的三级存储备份系统。为了在处理如此大的数据集时获得最高的效率和最高的灵活性，群集采用四处理器节点和每个节点8 GB的RAM。在全尺寸数据上进行测试可以消除在许多领域中由于对较小数据集进行采样而导致的错误。这些项目使新的研究成为可能，从分布式拒绝服务的分析，到生物医学图像的学习，再到并行断层合成。在这个设施上改进的软件工具包括信息检索、支持向量机（SVM）、聚类、网络模拟、核心外搜索、三维图像重建以及空间和时间数据库的方法。预计数据处理能力的提高将提供更快的周转速度，并使目前安装的技术无法获得的结果成为可能。

This project, acquiring a cluster to perform research on terabyte-scale problems in the areas of information retrieval, network traffic analysis, intrusion detection, and image processing for biomedicine, enables long term studies by providing access to users with terabytes of storage for extended periods of time. In this shared facility, investigators will cooperatively exploit a cluster containing 20 terabytes of disk storage. Additionally, a high performance terabyte disk subsystem, connected by fibre channel will serve as a higher speed cache for the large disk storage. To complete the memory storage, data migration will leverage using existing tertiary storage backup systems. In order to attain both the highest efficiency and the highest flexibility in processing such large datasets, the cluster employs quad-processor nodes and 8 GB of RAM per node. Testing on full size data eliminates errors caused by sampling smaller datasets in many areas. Enabling new research, the projects range from analysis of distributed denial of service, through learning for biomedical images, to parallel tomosynthesis. Software tools to be refined on this facility include methodologies for Information Retrieval, Support Vector Machines (SVM), Clustering, Network Simulation, Out-of-Core Search, 3-D Image Reconstruction, and Spatial and Temporal Databases. It is anticipated that the increase in data manipulation capabilities will provide much quicker turn-around and make possible results that are inaccessible with the currently installed technology.