Abstract

In this paper we present for the Bounded Disorder file organization a model that relates the secondary overflow bucket size to the primary bucket size of the data nodes (supposedly different from that of overflow) and the whole size of a data node, by considering the tree index. This result is obtained by minimizing the cost of inserting data into the structure and enables to compute the optimal overflow bucket size. Both the normal and partial expansion cases are considered and a model for the optimal relation of the parameters involved is developed.

Author Keywords: Bounded disorder; B-trees; Hashing