Oowards Multi-Resolution Database Management Systems

9312003 Silberschatz使用不同级别信息内容的数据对超大型数据库的性能至关重要，因为信息量较低(或分辨率较低)的数据较小，从而可大大减少输入/输出和通信费用。为了满足实时性能限制，建议采用以查询响应解析来换取更快性能的能力。只有支持以不同分辨率级别方便地检索数据的数据库管理系统(DBMS)才能充分利用多分辨率系统的这两个性能优势。构建了一个数据模型，该模型定义并支持查询响应的多分辨率和增量改进的概念，或“渐进式精化”。这个模型的基础是一种新的结构，称为“沙袋”，用于表示关于集合的不精确信息。开发了实现沙袋的实用、高效的算法，从而实现了多分辨率数据库管理系统。还开发了一种指定解析与性能的折衷的扩展查询语言，以及实现这种查询的机制。这项研究将为下一代数据库应用程序(如多媒体、科学、地理和其他超大型数据库)的信息处理挑战提供有效平衡性能和成本的解决方案。***

9312003 Silberschatz The use of data at different levels of information content is essential to the performance of very large databases because lower information content (or lower resolution) data is smaller and thus can significantly decrease I/O and communication costs. The ability to trade query response resolution for faster performance has been suggested as an approach to meeting real-time performance constraints. These two performance advantages of a multiresolution system can be exploited fully only by a database management system (DBMS) that supports the convenient retrieval of data at different levels of resolution. A data model that defines and supports the notions of multiresolution and incremental improvement of query responses, or "progressive refinement," is constructed. The basis of this model is a new construct, called the "sandbag," for representing imprecise information about a set. Practical, efficient algorithms that implement the sandbag, and thus enable a multiresolution DBMS, are developed. An extended query language that specifies the tradeoff of resolution for performance, and a mechanism for implementing such queries, are also developed. This research will provide solutions which effectively balance performance and cost for the information-processing challenges of next generation database applications such as multimedia, scientific, geographic, and other very large databases. ***