The Design And Implementation Of Remote Sensing Data Processing Grid Platform

With the development of the remote sensing technology, the update frequency of remote sensing information is becoming faster and faster. In remote sensing data processing, the dramatically increasing amount of remote sensing data and increasingly strict calculation accuracy requirements need huger computing power, which can hardly be met by the traditional calculation mode. Besides, there are super computers in only a few organizations. Meanwhile, in most organizations, there are large amounts of underutilized computing resources with a utilization rate of most desktop machines in less than 5%. More over, in some organizations, even the servers are not fully utilized. This is a huge computing and storage resource to be explored.The rapidly developing Grid technology brings possible solutions to the problem. As a newly emerging infrastructure, Grid will provide users with integrative service of information and application, which will fundamentally change our way of thinking and computing. The aim of Grid is to integrate the whole Internet as a super computer, realizing resources sharing and avoiding resource independence. In this way, there will be nothing different between the whole Internet and a PC in the users' perspective. The only difference is that the "PC" is super powerful, and you can obtain any resources you need such as computing resource, storage resource, bandwidth resource, software resource and data resource and so on, thus maximizing resource sharing.In view of the inevitable trend in the combination of remote sensing application and Grid technology, this paper studies the application of remote sensing data in Grid computing. The content is as follows:(1) The basic principle, concept and structure of grid computing is introduced. The shortcomings and defects of grid application are also pointed out. The current situation of grid application both at home and abroad is discussed. And the internationally well-known Grid projects are introduced.(2) The characteristics of remote sensing data and application are analyzed. The necessity and feasibility of the combination of remote sensing application and Grid technology is also pointed out and analyzed. And the complexity and time-consuming of remote sensing processing makes it necessary to use Grid computing to enhance its speed.(3) The Grid computing platforms are built based on the heterogeneous environment of high throughput computing system Condor by applying a dynamic centralized scheduling way, preliminarily realizing the load balance scheduling. A set of high-performanced aerosol remote sensing quantitative retrieval system is developed in this environment. The calculation time of Grid processing is reduced to 10 hours from 50 hours. In this way, the computing time of aerosol inversion is shortened and the requirement of real-time is also met.(4) A summary is made. The defects of this study are pointed out and the outlook in future researches is also prospected at the end of the thesis.