R-tree Updating Caching and Batch Processing Mechanism Based on Gird
Yansheng Lu · Jisuanji gongcheng · 2008
According to the relative stability of objects’ distribution and based on fixed grids,some leaf nodes representing the distribution of objects are selected as seed-nodes for new records to directly merge new records into seed-nodes without traversing R-tree.Some records which can not be merged into seed-nodes are chosen randomly as seed-records.Such records are grouped simply but effectively,so that they can be inserted into R-tree with the cost of inserting seed-nodes.The strategies pay attention to R-tree’s spatial clustering,and perform multiple inserting and deleting in one write-operation,decreasing the demand of writing and traversing R-tree.Experiments show that the strategies reduce the I/O cost of R-tree maintaining without affecting the query performance.