Supporting High Updates Disk-Based Index in Road Network

With the development of GPS and wireless techniques, more and more applications require maintaining the current position of moving objects on road network environment. It is the key stone that how to built an efficient index to supporting very high rates of spatial-index updates. Previous works in this domain assume that system holds enough main memory to buffer operations. This specialty enable these approaches don´t work as main memories is unavailable. To buffer the operations without main memory, we propose a new R*-tree-based indexing technique, called disk-based R*-tree (DBR*tree for short), which buffers update operations in disk page as well as grouping them to reduce disk I/O. DBR*-tree employs R*-tree to index the edges on the network (for example the road), not moving object, which eliminates the expensive costs caused by frequent changes in R*-tree structure while indexing moving objects by R*-tree. Finally, we present performance analysis and experimental results. And both of them show the proposed technique holds high performance on update and query operations.