Online Data Migration

在线数据迁移

• 批准号: 0328393
• 负责人: Betty Salzberg
• 金额: $ 20.23万
• 依托单位: Northeastern University
• 依托单位国家: 美国
• 项目类别: Continuing grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-09-01 至 2007-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0328393&HistoricalAwards=false
• 关键词: Online; Data; Migration

Online Data MigrationCurrent demand for scalability in databases leads to an architecture of loosely-coupled computer nodes which can be expanded or contracted at will. Demand for 24/7 availability leads to requirements for online reorganization. To satisfy these demands, this project develops algorithms for migrating data from one node to another in a distributed system while still allowing users to access the data. First, several schemes for deciding which data to move are considered. While the goal is load balancing, performance of the migration algorithm itself is also a consideration. Algorithms must move only a small amount of the data, not reassign everything to a new node. Experiments are run to measure the "goodness" (coefficient of variance) of load balancing as well as the cost of migration. Second, index maintenance during migration is considered. When only a relatively small amount of data is moved, creating new indexes is more expensive than merging indexes or index entries of moved data into already existing indexes at destination nodes. This merging process must also be online so that users have access to all the data almost all of the time. Experiments measure both efficiency and concurrency for index merging algorithms. Broader impacts of this research include (1) educating the future educators of our technological workforce, (2) enabling efficient online reorganization of massive databases, such as occurs when government agencies are combined and (3) aiding the database industry in satisfying customer demand for scalability and 24/7 availability. More information about this project is available at http://www.ccs.neu.edu/research/dblab/

在线数据迁移当前对数据库可伸缩性的需求导致了松散耦合的计算机节点体系结构，这些节点可以随意扩展或收缩。对24/7可用性的需求导致了对在线重组的需求。为了满足这些需求，该项目开发了将数据从分布式系统中的一个节点迁移到另一个节点的算法，同时仍然允许用户访问数据。首先，考虑几种用于决定移动哪些数据的方案。虽然目标是负载平衡，但迁移算法本身的性能也是一个考虑因素。算法必须只移动少量数据，而不是将所有数据重新分配给新节点。运行实验来测量负载平衡的“善良”（方差系数）以及迁移的成本。 其次，考虑了迁移过程中的索引维护。当只移动相对少量的数据时，创建新索引的开销比将索引或已移动数据的索引条目合并到目标节点上的现有索引中要大。这个合并过程也必须是在线的，这样用户几乎可以随时访问所有数据。实验测量索引合并算法的效率和并发性。这项研究的更广泛的影响包括：（1）教育我们技术劳动力的未来教育者，（2）实现对大型数据库的有效在线重组，例如当政府机构合并时，以及（3）帮助数据库行业满足客户对可扩展性和24/7可用性的需求。 有关此项目的更多信息，请访问http://www.ccs.neu.edu/research/dblab/