FAUVE: Specification and Implementation of Delayed Consistency for Replicated Data in Federated Databases

FAUVE：联邦数据库中复制数据延迟一致性的规范和实现

• 批准号: 9007753
• 负责人: Jeffrey Ullman
• 金额: $ 14.58万
• 依托单位: Stanford University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1990
• 资助国家: 美国
• 起止时间: 1990-09-15 至 1993-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9007753&HistoricalAwards=false
• 关键词: FAUVE; Specification; Implementation; Delayed; Consistency

Practical distributed database systems operate without rigid centralized control. Such systems are considered to be federated distributed systems, and attractive for involving many sites, dispersed authority, and differences in requirements of level of data abstraction and of all data being-up-to-date. The interaction via communication networks, the partial limited availability of network nodes, delays in local data acquisition, and the actual delay and cost of computations to derive data for remote sites all conspire to defeat simple approaches to network- wide data consistency. The problems of lack of synchrony among the nodes have been recognized piecemeal and many methods and heuristics to manage data in this environment have been proposed. We have a simple model to permit designers and users to specify the extent to which consistency can be temporarily relaxed. The FAUVE research classifies existing algorithms and heuristics, identifies conflicts in assumptions made, and recognizes complementary approaches. It provides a framework to focus on situations now not adequately dealt with. This framework provides the users with specifications, and provides database operations on federated databases with a measure of success or failure. Such a measure is essential for the scientific progress in this subtopic of modern computer systems research.

实际的分布式数据库系统运行时没有严格的集中控制。这样的系统被认为是联邦分布式系统，因为涉及许多站点、分散的权限以及数据抽象级别和所有数据都是最新的不同需求而具有吸引力。通过通信网络的交互，网络节点的部分有限可用性，本地数据采集的延迟，以及为远程站点导出数据的实际延迟和计算成本，所有这些共同挫败了实现网络范围数据一致性的简单方法。节点间缺乏同步的问题已经逐渐被认识到，并提出了许多在这种环境下管理数据的方法和启发式方法。我们有一个简单的模型，允许设计师和用户指定可以暂时放松一致性的程度。FAUVE研究对现有算法和启发式进行分类，识别假设中的冲突，并识别互补方法。它提供了一个框架，把重点放在目前没有得到充分处理的局势上。该框架为用户提供规范，并为联邦数据库上的数据库操作提供成功或失败的度量。这种措施对于现代计算机系统研究的这一分支的科学进步是必不可少的。