The Application Of Parallel Algorithm In Meteorological Satellite Ground Application System

FY-3 meteorological satellite (Abbreviations: FY-3) ground application system is a supporting project of new generation polar-orbiting weather satellite-FY-3 weather satellite. It is a very important one of the five system of FY-3 meteorological satellite project, the date-fetching and processing system building for receiving, processing, storing, distributing, and applying the satellite products. The business operation platform is a high-performance heterogeneous cluster system, with complex of business relation, large number of parallel tasks, strong requirements of real-time, requirements of high-performance and big throughput of data, and the scheduling system is the mainly carrier to realizing the functions of the business. Designing a effective parallel algorithm to balancing the workload of the system, sufficiently using the resource of the cluster, increasing the efficiency of date parallel-processing, thereby achieving the high real-time requirement of meteorological satellite date processing, making the system running efficiently, steadily and automatically, is the emphases of this paper.At first, the paper did a systemic research in the key technology of job-scheduling system and the representative cluster job management system, and make contrast among them.The emphases of this paper are chapter three and chapter four. Based on the high performance of the SMP cluster, considering the characteristic of FY-3 satellite data processing task, the paper discusses the target system's job-scheduling strategy which is the flow-job hierarchical scheduling strategy. In the data flow level, sub-data-flow processing parallel scheduling strategy is adopted; in the job and job-flow level, four-level job-scheduling strategy based on XML is customized. Basing on the analyse of the feature of date-processing of the business system, drawing on the experience of Ants Algorithm, Reserving Resources, Priority strategy and virtual server application concept, the paper puts forward a parallel-scheduling processing algorithm——pheromone with dynamic weight algorithm(pdw) based on performance forecast and link weight calculated, and applies it in the business system.Chapter five of the paper shows the result of pdw algorithm load balance test and applying effect, the pdw algorithm could balance the system workload effectively, it also could increasing the efficiency of date-processing.At last, author summarizes the shortages of this job-scheduling system, and then points out the Direction of the job-scheduling system's overall performance to be further enhanced.