A Raster Map Tiling System Based On MapReduce

In recent years, the rapid development of geographic information technologies and continuous improvement of remote sensing instruments resolution cause an unceasing extension of map applications and rapid growth in raster map data. Current Web GIS usually tiles raster map data in advance, and then organizes the data in the form of tile pyramid. If a client accesses Web GIS to obtain the map data, the server will respond with the tile data tiled in advance. Traditional standalone map tiling systems suffer the restriction of computation and storage capacity in a single node, thus forming the bottleneck of massive raster map data processing technologies gradually. Accelerating the tiling speed of massive raster map data is of great significance for improving the efficiency of Web GIS.Cloud computing technology can manage data and computing tasks on a large number of computer nodes, and has the characteristics of high reliability, high expansibility and so on. Cloud computing technology provides unprecedented capacity in computation and storage, and a new foundation platform for accelerating the tiling speed of massive raster map data. This paper solves the tiling problem of massive raster map by building the raster map tiling system based on Map Reduce, a distributed computing framework. The main research contents and innovations of the paper are as follows:1. Key technologies of the existing distributed map tiling systems have been studied. Firstly, the weakness of these systems is proposed. Secondly, according to the tiling requirements of massive raster map data, a raster map tiling system based on Map Reduce is designed.2. The redundant backup mechanisms of HDFS have been studied. A splitting algorithm of map files is proposed. The network data transmission during map tiling has been reduced by using “local computing” mechanism of Map Reduce. In addition, the dependence of tiling system on the network has been reduced by running tiling tasks on the node which actually stores data.3. The building schemes of tile pyramid have been studied. Firstly, according to the model structure of tile pyramid, a building schema using tile cache of tile pyramid is designed. The building speed of tile pyramid has been accelerated by reducing the amount of computation during tile scaling. Additionly the memory consumption has been reduced by using the arrangement rule of tiles in the pyramid.4. The parallel building schemas of tile pyramid on massive raster map data have been studied. Firstly, according to the characteristics of Map Reduce application, an iterative building schema of tile pyramid is proposed. The parallel building of tile pyramid has been achieved by distributing formatted block files to a large number of nodes. Secondly, the iterative tiling has been accomplished by merging tile data at the bottom of the pyramid as the input data of the next round. Thirdly, the tiling speed of massive raster map has been accelerated by using the powerful computation and storage capacity of cloud computing.This paper implements the raster map tiling system according to the above research results. A time related performance testing is constructed to validate the feasibility and efficiency of the design. The testing result shows that the tiling time will decrease linearly with the increment of nodes number in cluster. This paper improves the tiling speed and solves the tiling problem of massive raster map data by using cloud computing technology.