Abstract: The application of high performance computing cluster technology to remote sensing data processing is one solution to meet the requirements of real-or near-real-time and high-accuracy processing capabilities. Job schedule and management is an important component of cluster which is designed to improve the throughput rate and resource utilization rate of system. Low degree of automation, blindness of resource selection and short of feasibility is unavoidable in common schedule system. Aiming at remote sensing data cluster processing, based on open sources Portable Batch System（PBS）, it develops a twin-stage scheduling strategy for resources assigned optimally and automatically. With the layered structure and service-oriented programming method, the job scheduling and management system for cluster-based remote sensing data processing system is C/S pattern which is communicated with ICE（Internet Communications Engine） middleware by XML（Extensible Markup Language） format data transmitting.

Key words: twin-stage scheduling strategy, optimal resources assignment, data processing of remote sensing, cluster

摘要： 高性能遥感数据集群处理技术提高了遥感数据处理的速度和精度。作业调度管理是集群处理系统的核心组成部分，有助于提高系统的吞吐率和资源使用率。常用调度管理系统通用性强，但针对性差，自动化程度低，资源使用存在盲目性。针对遥感数据集群处理，基于集群作业调度管理系统Open PBS，提出一种最优资源自动分配的“双级调度策略”；应用分层架构和面向服务的软件开发思想，采用ICE中间件的网络通信技术，使用XML格式作为数据交换模式，研究并实现C/S（Client/Server）架构的灵活、友好、高效的遥感数据集群处理作业调度管理系统。

关键词: 双级调度策略, 最优资源分配, 遥感数据处理, 集群