Improving the Storage Utilization of 0-Complete Trees as Index Structures

Abstract

A compact 0-complete tree is an indexing mechanism that can be effectively used for large databases with long and variable-size keys. Compared to B-trees, compact 0-complete trees eliminate search values from secondary indices altogether. They are replaced with small surrogates whose typical eight-bit length will be adequate for most practical key lengths. Hence the secondary indices are simply hierarchical collections of (surrogate, pointer)-pairs. However, the overall storage performance of a compact 0-complete tree structure suffers from its dependency on the distribution of keys. We present a structural modification of 0-complete trees coupled with retrieval and maintenance algorithms that improve their storage utilization by reducing page-underfilling. This is achieved by introducing the notion of a sequence set that is analogous to that of $\text {B}^{+}$-trees.