| Cluster system based on Linux has become one of the most popular high performance computation platforms whose ratio occupancy in the high performance computer is more and more bigger,and brings the chance and challenges for the parallel processing and distributed computing of spatial information.Spatial data characteristics such as the super and big capacity, distribution extensively and diversity of type,and combinational problems such as dynamic, non-linear, many dimensionses of time and space which brings at the time of the spatial data processing bring the computation complexity for analyzing and processing the spatial data in the network environment.Therefore,how to make cluster powerful in computing capacity and throughput is the problem that the system administrator and the program members consider together. The experience expresses that it is absolutely necessary to tailor an efficient task scheduling mechanism. It is easier to implement the mechanism than develop parallel algorithms. Job management system is a subsystem of cluster to carry out the job scheduling and management. Designing job management system oriented to the spatial data processing for cluster is to tailor the homologous scheduling algorithm according to uses' particular requirement and job loading characteristics of application domain of GIS.the further task is to optimize various cluster computation resources such as the CPU, memory, hardware, network bandwidth and the application program, etc,increase the speed of responding when process and analyze the spatial data, finally improve the whole performance of system in GIS applications.The Job Management System usually consistes of resource management system and job scheduling system. The functions of Resource Management System is to collect the attribute and status of various resources, all requests submitted by users and execute jobs. Job Scheduling System creates queue priority listings according to the collected data and the appointed scheduling strategy, by which Resource Management System starts to execute jobs. Thus, Job Management System can control the site, and make site's task strategy map up scheduling behavior, accordingly increasingly improve the running efficiency for the whole cluster. Therefore, how to choose a Job Management System for cluster system and tailor scheduling algorithms for the spatial data application is a key in cluster-based Job Management System oriented to spatial data processing.This paper takes Portable Batch System(PBS) as the prototype. Based on analyzing the PBS system structure, the resource management machine(Torque) and job scheduling machine (Maui) are integrated and taken as the job management platform of proceeding the GIS high performance applications.In discussing the spatial data and it's application characteristics, the comprehensive algorithm based on BackFill algorithm is chosen as the scheduling algorithm of processing the spatial data.Finally, the article designs and realizes the prototype system oriented to the spatial data and verifies the feasibility of this kind of method by the test. It can neither let the coarse grain job waiting, and nor affect the efficiency of the fine grain job execution. Thus it raises the whole performance of processing spatial data in cluster. |
