Object Clustering Heuristics and Algorithms in Next Generation Database Management Systems

点击翻译按钮获取中文摘要

Traditional (e.g., relational) database systems cluster data on disks in order to minimize the cost of accessing information; such mechanisms are simplistic and are designed before the system is put into use - and are never changed. But new applications are making bigger demands on database systems. An example is aircraft engineering, where multiple designers simultaneously perform ever-changing manipulations on complex integrations of electronic and mechanical engineering components. This is in contrast to data processing, where highly-predictable data manipulations are typically performed. This has caused data clustering to emerge as a critical area demanding newer and more powerful algorithmic results. The goal of the self-adaptive clustering project is to consider the wide variety of factors that come into play when one considers clustering algorithms which are not constructed in advance and can adapt themselves to changing, unusual database access patterns. This will lead to the development of a theory of self-adaptive clustering, which will be of great significance to a wide variety of database users, in particular CAD/CAM, software, printed circuit board, and VLSI designers. This will provide a spectrum of software algorithms which will transform overly-expensive applications into cost-effective systems.