External Memory Data Structures for 3 and 4-Sided Queries

In this paper, we experimentally evaluate the relative performance of 3 external memory query structures. The R and Priority R (PR) Trees are compared using 4-sided rectangle queries in R. For 3-sided point queries in R, we compare the two structures mentioned above to the External Priority Search (EPS) Tree. While the PR and EPS Trees possess optimal I/O bounds, they are largely theoretical works. The R-tree, on the other hand, is a simpler heuristic data structure that is commonly used in practise but is inefficient in the worst case. After comparing wall times and I/O efficiency of queries, updates and bulk loads of the three structures on a variety of simulated and real data, we find the R-Tree variants to generally outperform the EPS and PR trees in R. In higher dimensions, however, the PR-Tree is most efficient.