A New Operation for Two R-trees to Efficiently Migrate Spatial Objects

Abstract

In the past, several papers discussed the access of individual object in R-tree, but rare mentioned how to access several objects simultaneously. In this paper, we propose a new operation, called spatial-migration, for R-tree. The function of this operation is to combine one group of objects into another group of objects according to a special relationship between the two groups of objects. That is, more than one spatial object in one R-tree is migrated to another R-tree at the same time. If objects are migrated one by one, several R-tree nodes may overflow or underflow repeatedly. The database performance may decrease because the R-tree may be reconstructed again and again. When many single-object insertions/deletions are replaced by a single multiple objects insertion/deletion, many redundant node-splits and/or MBR-adjustments can be omitted. When a node overflows due to the insertion of many objects, we once generate enough nodes to contain all the objects inserted into the node. Each node at most has only one node-split and/or MBR-adjustment. Therefore, the proposed spatial-migration operation can efficiently migrate objects between two R-trees without effecting database performance much.