GBI: A Generalized R-Tree Bulk-Insertion Strategy

A lot of recent work has studied strategies related to bulk loading of large data sets into multidimensional index structures In this paper we address the problem of bulk insertions into existing index struc tures with particular focus on R trees which are an important class of index structures used widely in commercial database systems We pro pose a new technique which as opposed to the current technique of inserting data one by one bulk inserts entire new incoming datasets into an active R tree This technique called GBI for Generalized Bulk In sertion partitions the new datasets into sets of clusters and outliers constructs an R tree small tree from each cluster identi es and pre pares suitable locations in the original R tree large tree for insertion and lastly performs the insertions of the small trees and the outliers into the large tree in bulk Our experimental studies demonstrate that GBI does especially well over better than the existing technique for randomly located data as well as for real datasets that contain few natu ral clusters while also consistently outperforming the alternate technique in all other circumstances Index Terms Bulk insertion Bulk loading Clustering R Tree Index Struc tures Query Performance This work was supported in part by the University of Michigan ITS Research Cen ter of Excellence grant DTFH X Sub sponsored by the U S Dept of Transportation and by the Michigan Dept of Transportation Dr Rundensteiner thanks IBM for the Corporate IBM partnership award and Li Chen thanks IBM for the Corporate IBM fellowship as well as mentoring from the IBM Toronto Labora