The Performance of the hB - pi Tree

The hB-pi tree is a multi-attribute index method which combines the hB tree, a multi-attribute access structure and the II-tree, an abstract index with efficient concurrency and recovery methods. The hB-pi tree is a balanced tree and adapts well to different data distributions. Furthermore, the hB-pi tree inherits the efficient concurrency and recovery methods from the II tree. This paper provides the implementation details of the hB-pi tree and explores the performance of different hB-pi tree variants. The paper begins by illustrating the experimental results on real and synthetic point data sets. The hB-pi tree cannot directly index the non-point objects and different conversion methods are applied to the hB-pi tree to index the non-point data. The experimental results on real and synthetic non-point data sets are also provided. The experimental results are analyzed in order to proposal the future research direction.