Empirical evaluation of strategies to process range queries of numeric sequences in batch-mode

Tree structures are largely used to index and search sequences on secondary memory. In some situations, many range queries are processed almost simultaneously and the resulting number of disk accesses can be high. In order to reduce the number of disk accesses, similar sequences can be grouped and spanned as a single query. A simple strategy is to unify all sequences into a single group. However, other strategies for grouping sequences can also be used. In this paper, we present and empirical evaluation of 5 common grouping strategies for R-trees andM -trees. Our results indicate that for inputs modelled as a random walk distribution the overall best implemented strategy for grouping queries is indeed the one unifying all queries in a single group.