The efficiently using complex objects becomes more and more important in Object-Oriented Database. In this paper, we propose an adaptive clustering algorithm that can cluster complex objects off-line, and suggest the physical positions in disk. By the way, we also examine the problem that clustering algorithm with buffer strategy supported. It is shown that we must adjust the CSDG (Complex Structure Directed Graph) Weighting Factor with the object access frequency consideration both in memory and disk. Consequently, the simulation results show that our clustering algorithm performs better compared with pure leveled clustering algorithm and suits for various kinds of database environment.
展开▼