GridDB-Lite: Database Support for Data-Driven Scientific Applications in the Grid

该项目的目的是开发一个中间件框架，使查询能够有效地执行，以便在分布式环境中从大型科学数据集中提取感兴趣的数据。该项目将开发和评估一个中间件基础设施，Grid database - lite (GridDB-Lite)，以支持网格中的以下基本数据库操作：从分布在不同存储系统中的数据集中选择感兴趣的数据；并将数据从存储系统传输到计算节点进行处理。该架构将利用两个已开发的框架：Active Data Repository和DataCutter。该项目还将：设计一个框架，将用于查询科学数据集的对象关系模型纳入其中；开发高效算法；检查分布式环境中大型数据集的索引和聚类方法；并开发和评估算法，以有效地将数据集的选定子集从分布式存储集群传输到目标计算集群上的处理器存储器或磁盘。更广泛的影响和智力价值-本研究项目提出解决与大规模数据分析的关键步骤有关的问题，即从大型分布式数据集中提取感兴趣的数据。随着数据集规模的持续增长，这一步的有效解决方案对于在网格中大规模部署数据分析将变得越来越重要。该项目还将对教学和人力资源发展产生重大影响。作为生物医学信息系课程的一部分，申请者将开设一门关于大规模数据管理和操作的课程（BMI731）。本课程使生物医学信息学的学生了解网格技术，并使计算机科学家能够发现新的数据密集型网格应用。本课程的学生将获得GridDB-Lite的经验，并将接触到应用程序开发。本项目的大部分资金将用于支持博士研究生，培养他们在计算机科学领域做出长期研究贡献。

The purpose of this project is to develop a middleware framework to enable efficient execution of queries for extracting the data of interest from large scientific datasets in a distributed environment. The project will develop and evaluate a middleware infrastructure, Grid Database-Lite (GridDB-Lite), to support the following basic database operations in the Grid: selection of the data of interest from datasets distributed among disparate storage systems; and transfer of data from storage systems to compute nodes for processing. The architecture will leverage two developed frameworks: Active Data Repository and DataCutter. The project will also: design a framework to incorporate an object-relational model for querying scientific datasets; develop algorithms for efficient; examine methods for indexing and declustering of large datasets in a distributed environment; and develop and evaluate algorithms for efficient transfer of selected subsets of datasets from distributed storage clusters to processor memories or disks on destination compute clusters.Broader impact and Intellectual merit - This research project proposes to address issues that pertain to a key step in large scale data analysis, which is the extraction of the data of interest from large, distributed datasets. As dataset sizes continue to grow, an efficient solution to this step will be increasingly important to wide-scale deployment of data analysis in the Grid. This project will also have a significant impact on teaching and human resource development. As part of the curriculum in the Biomedical Informatics Department, the proposers will offer a course (BMI731) on large-scale data management and manipulation. This course exposes students in biomedical informatics to grid technologies, and enables computer scientists to discover new data-intensive grid applications. Students in this course will gain experience with GridDB-Lite and will be exposed to application development. A majority of funds in this project will be used to support Ph.D students, and to train them to make long-term research contributions in computer science.