排序方式： 共有2条查询结果，搜索用时 0 毫秒
1
1.
通过分析与研究建立资源管理和作业动态分配ProRMJS模型,解决了中国地震灾害仿真网格试验系统（CEDAGrid）中网格科学计算平台在初期建设中存在的一些资源管理方面的问题. 针对网格科学计算平台默认各个计算节点均能提供计算服务的问题,ProRMJS通过ldquo;计算池rdquo;这一虚拟组织为调度器分发作业提供支撑;而调度器则根据计算池中各个节点的计算能力和状态,动态分配任务,这样就保证了计算平台的稳定性. 针对默认各个计算节点均能完成任务的问题, ProRMJS通过监控各个节点所负担作业的运行进度和设定作业时间阈值来管理作业运行. 对于各个节点计算能力大小的区别, ProRMJS通过对各个节点计算性能的权值量化,实行按ldquo;需rdquo;分配作业. 对各个节点计算性能的权值量化, ProRMJS是通过综合考虑各个因素对机器计算能力大小的影响后通过加权计算的方式完成的,从而提高了整个计算平台的工作效率. 最后通过实现地震应力触发科学计算实例验证了方案的有效性,为网格技术在地震领域内的进一步应用作了有意义的探索. 相似文献
2.
Grid technique is taken as the third generation internet technology and resource management is the core of it. Aiming at the problems of resource management of CEDAGrid （China Earthquake Disaster Alleviation and Simulation Grid） in its preliminary construction, this paper presents a resource management and job scheduling model： ProRMJS to solve these problems. For platform supposed agreeably each computing node can provide computation service, ProRMJS uses ＂computation pool＂ to support scheduler, and then the scheduler allocates jobs dynamically according to computing capability and status of each node to ensure the stability of the platform. At the same time, ProRMJS monitors the status of job on each node and sets a time threshold to manage the job scheduling. By estimating the computing capability of each node, ProRMJS allocates jobs on demand to solve the problem of supposing each node can finish the job acquiescently. When calculating the computing capability of each node, ProRMJS allows for the various factors that affect the computing capability and then the efficiency of the platform is improved. Finally the validity of the model is verified by an example. 相似文献
1