Research On Distributed And Parallel Spatial Index Mechanism

Nowadays, with the information technology has been more and more popular in many application, spatial databases are being applied widely in many applications such as GIS, CAD, Robot, Compute Geometry, Physic Image, Multimedia System and Computer Vision and so on. Along with the concepts of Digit Earth, Digit-City and Digit-River being put forward and applied, the higher efficiency of storing and processing spatial data is demanded. Spatial index is vital technique for improving the performance of spatial databases. It affects the storage efficiency and spatial retrieve performance of spatial data directly. Studying spatial index techniques and investigating better spatial data index mechanisms have been a focus of research in the computer circles and other application field.The dissertation starts from the introduction of classic spatial database index technology in the commercial spatial database system. First of all, the basic theory and main methods of all all kinds spatial index technology are described, and their advantages and disadvantages which determine their applied occasions are also pointed out in the following. Moreover, the basic theory of distributed system and parallel system and the characteristic of distributed spatial database and the structure of distributed and parallel spatial database are introduced. Once more, a new structure of distributed parallel index structure-DPR tree are constructed.It is based on parallel Master-Client R tree, and make some change to adapt to the distributed environment, at the same time, introducing HCSDP division technology, and applying it to distributed environment, and the DPR tree's sub-tree use the improving R-tree—R*Q tree. DPR tree is suitable for the environment of massive data or high ratio spatial data, and the query efficiency is highly effective.While the DPR tree structure is analyzed and test, it is compared to the DPslR+ tree, because their index structure is so similar in research thought. By means of the average inquiry time is analyzed and practical test show that the DPR tree index structure are higher than DPslR+ tree in inquiry performance. Besides, The DPR tree is smaller in required memory and well-distributed in data storage capacity and lower in disks access rate.