论文部分内容阅读
随着并行空间计算任务的不断增多,传统的MPI服务器集群容易出现排队时间长、拒绝服务甚至系统瘫痪等情况。利用虚拟化、作业调度等技术构建云计算平台上的MPI虚拟集群可以提升MPI服务性能,从整体上缩短排队等待的时间,从而使得服务质量QoS(Quality of Service)得到保证。通过在OpenStack上部署MPI虚拟服务器集群、利用Torque实现MPI作业的调度管理的基础上,使用DEM(Digital Elevation Model)等高线生成算法组成MPI作业队列,对传统MPI物理集群与MPI虚拟集群进行性能对比分析,结果显示了云计算平台上MPI并行环境在面对大量任务作业时的优势。
With the increasing parallel computing tasks in space, traditional MPI server clusters are prone to queuing a long time, denial of service and even system paralysis. Constructing the MPI virtual cluster on the cloud computing platform using virtualization and job scheduling technologies can improve the performance of the MPI service and shorten the waiting time of queuing as a whole, so that Quality of Service (QoS) can be guaranteed. Based on the deployment of MPI virtual server cluster on OpenStack and the scheduling management of MPI jobs using Torque, MPI job queues are constructed by using DLE (Height Elevation Model) contour generation algorithm and the performance of traditional MPI physical clusters and MPI virtual clusters Comparative analysis shows that the MPI parallel environment on the cloud computing platform has the advantage of facing a large number of tasks.