ITR: Development of a Data Engine for Grid-Enabled Analytical Computing

ITR：开发用于支持网格的分析计算的数据引擎

• 批准号: 0635365
• 负责人: Ratko Orlandic
• 金额: $ 11.66万
• 依托单位: University of Illinois at Springfield
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-08-16 至 2007-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0635365&HistoricalAwards=false
• 关键词: ITR; Development; Data; Engine; Grid

The objective of this project is to support the Data Grid infrastructure by developing new and original techniques for efficient storage, retrieval, and analysis of complex scientific data. The main focus is on the development of a highly scalable data engine geared toward the needs of analytical computing in Data Grid environments. However, keys to the realization of the project are the advances in the areas of indexing and clustering data in multi-dimensional spaces.While the main goal of analytical computing in Data Grid environments is to facilitate hypothesis formulation or to test the validity of a postulated scientific model, its primary method is usually that of data clustering. Since typical analytical tasks also rely on ad-hoc data exploration, any data engine for Grid-enabled analytical computing must support an integrated set of different retrieval and clustering techniques. The data engine developed in this project will feature: an efficient and scalable indexing technique for data in high-dimensional spaces, which will include a practical solution for handling data with missing information; a new and original access method for similarity searching in multi-dimensional spaces; and an original technique for clustering large volumes of multi-dimensional data, which will require no dimensionality reduction.

该项目的目的是通过开发新的和原始的技术来支持数据网格基础架构，以有效地存储，检索和分析复杂的科学数据。主要重点是开发高度可扩展的数据引擎，该引擎针对数据网格环境中的分析计算需求。但是，实现项目的关键是多维空间中索引和聚类数据领域的进步。尽管数据网格环境中分析计算的主要目标是促进假设公式或测试假定的科学模型的有效性，但其主要方法通常是数据群集的主要方法。由于典型的分析任务也依赖于临时数据探索，因此任何用于网格分析计算的数据引擎都必须支持一组集成的不同检索和聚类技术。该项目中开发的数据引擎将具有：用于高维空间中数据的高效且可扩展的索引技术，其中包括用于处理数据中缺少信息的实用解决方案；在多维空间中搜索相似性搜索的一种新的和原始的访问方法；以及一种用于聚集大量多维数据的原始技术，这将不需要降低维度。