Abstract: In parallel spatial database, it is necessary to make the spatial data set cluster in each node, because it can improve the efficiency of parallel database query. The partition approach of Oracle Spatial is based on grid. It only consi deres data sets in each node are a balanced division, without taking into account the topological characteristics of these data. In order to improve the problem, this paper presented a new spatial data partition approach which is based on K-means clustering algorithm. Experiments show that the method greatly improves the spatial data retrieval and ctuery efficiency in parallel.

Key words: K-means algorithm,Cluster,Data partition,Distributed & parallel computing