CISE Reasearch Instrumentation: A Data-Intensive DistributedComputing System

9617465 Solworth, Jon A. Lobo， Jorge University of Illinois at Chicago CISE研究仪器：数据密集型分布式计算系统随着越来越大的数据集（包括结构化和非结构化）的创建和数字化，需要新的技术来更好地存储、组织和访问这些数据。这是具有挑战性的，因为大容量存储子系统的性能相对于处理器性能增长缓慢，而且数据通常分布在使用高速网络通信的计算机上。因此，需要新的存储优化技术来提高性能和可用性。这项研究仪器补助金有助于获得一个由工作站组成的系统，该系统具有非常大的实验磁盘和磁带容量，通过非常高速的网络进行通信，以授权以下项目的研究：-持久对象存储系统-动态分布式复制实验-按内容检索图像和视频系统-分布式数据库处理系统对现实数据密集型应用和系统的实验评估将因此成为可能。该设备将用于构建基于内容的图片视频检索系统，实现大型视频档案的按特征查询。分布式演绎数据库系统最初旨在回答蛋白质数据库查询，实现基于规则的数据分析。这些应用程序以原型形式存在，现在将扩展到实际大小的系统，从而更好地理解性能和语义。与这些应用相辅相成的是构建大规模对象存储和分布式数据复制的系统研究。这一系统研究将有助于构建性能更高、更灵活的存储，同时减轻编程负担。

9617465 Solworth, Jon A. Lobo, Jorge University of Illinois at Chicago CISE Research Instrumentation: A Data-Intensive Distributed Computing System As ever larger data sets--both structured and unstructured--are created and digitized, new techniques are need to better store, organize, and access this data. This is challenging since mass storage subsystem performance has increased slowly relative to processor performance and because data is usually distributed over computers which communicate using high-speed networks. Hence, new storage optimization techniques are needed for both performance and availability. This research instrumentation grant facilitates acquiring a system consisting of workstations with very large experimental disk and tape capacity, communicating over very high speed networks to empowered research in the following projects: -Persistent Object Store System- Dynamically Distributed Replication Experiments-Picture and Video Retrieval by Contents System-Distributed Database Processing System Experimental evaluation of realistic data-intensive applications and systems will thus be enabled. The equipment will be used to construct a picture and video retrieval by contents application enabling large video archives to be queried by characteristics. A distributed deductive database system targeted initially at answering protein data bank queries enables rule-based data analysis. These applications exist in prototype form and will now be scaled to realistic sized systems, enabling better understanding of performance and semantics. Complementing these applications is system research on constructing large scale object stores and distributed replication of data. This systems research would enable the construction of higher performance, more flexible stores while easing the programming burden.