Alternatives to the B⁺- tree

Abstract

For many years, the b+ tree has been the file structure of choice for applications requiring both sequential and direct access. But two file structures developed recently for similar applications, bounded disorder file organization and adaptive hashing, may offer improvements. Are they better? In what ways? When should they be used? To answer these and related questions, this paper summarizes the performance characteristics, implementation aspects, and use considerations for these three file structures. The primary variation among the three techniques is the form of the index structure. Results from an investigation comparing the three methods are presented; parameters varied in the study were the distribution of the record keys, file size, and data node size. For many circumstances, one of the newer techniques is preferable to the b⁺-tree.