An empirical comparison of B-trees, compact B-trees and multiway trees

Abstract

It is well-known that the B-tree data structure yields excellent worst-case search costs and for that reason is widely employed in the organization of external files and in the implementation of data bases. In this paper, we examine general B-trees empirically and compare them with a less restrictive structure, the general multiway tree, and a more restrictive structure, the compact B-tree. We compare search costs, insertion costs, and space costs of these three structures for both small and large orders and indicate their relative utility for large and small data sets. Although there are cases when general multiway trees are more effective than B-trees, this is not the case for most practical situations. Compact B-trees are also shown to degrade rapidly in the presence of insertions and are therefore only useful for static data sets.