Presidential Young Investigator Award

该项目涉及数据库研究的三个领域：几何数据、文本和多媒体归档系统以及多磁盘文件分配。为了寻找管理几何数据的最佳方法，将对几种有前景的方法进行分析和实现以进行比较。重点介绍空间填充曲线和分形，如Peano曲线、Hilbert曲线、基于格雷码的曲线，因为它们可以产生良好的距离保持映射。第二个项目涉及开发多媒体文件归档系统。该系统的新颖特点是，它将使用签名文件进行文本检索，它将使用光盘进行存储，并且它将允许在文档中使用静态和动态图像。将研究将文件存储在多个磁盘上以最大限度地提高并行度的分散技术。这个目标与第一个项目相反。问题是将不同的记录组合在同一张磁盘上。已经开发出一种启发式方法，它使用信息论和纠错码，因为这类码的码字故意不相似。初步实验表明，该启发式算法的性能始终优于传统的自组织启发式算法。

This project is concerned with three areas of database research; geometric data, text and multimedia filing sytems and multiple disk file allocations. In search of the best method to manage geometric data, several promising methods will be analyzed and implemented for comparison. Emphasis will be given to space filling curves and fractals, such as the Peano curve, the Hilbert curve, curves based on Gray codes, because they lead to good distance preserving mappings. The second project involves the development of a multimedia document filing system. The novel characteristics of the system is that it will use signature files for text retrieval, it will use optical disks for storage and it will allow static and animated images in the documents. Declustering techniques that store a file on multiple disks to maximize the parallellism will be studied. This goal is the inverse of the first project. The problem is to group unsimilar records on the same disk. A heuristic that uses information theory and error correcting codes, because the codewords of such codes are deliberately unsimilar has been developed. Preliminary experiments show that this heuristic consistently outperforms older ad hoc heuristics.