A Novel GPU Resources Management and Scheduling System based on Virtual Machines
With the capabilities of hardware resources increasing while prices also continuing to decline,virtual machine technologies have become more and more popular in industrial and academic areas to utilize computer resources fully while completed isolating the users.Virtual machine technologies o er lots of bene ts such as users isolation,server consolidation and live migration.However,due to the overhead incurred by indirect access to physical resources such as GPU,IO devices,VM technologies have not been widely used in high performance computing area.Since GPGPU (General Purpose Graphics Processing Unit) computing solution for virtual machines being proposed,GPU can be used to accelerate applications in virtual machines (VMs),which marks high performance computing (HPC) applications running on GPU can be ported to VM environment.However,there are some di erences between GPU cluster built on virtual machine environment and real physical environment.A novel GPU resources management system built on VMs,called VMGPURMS,is presented to use VM as computing node,and enable users run the jobs in the whole virtual machine environment transparently.Also a novel GPU cluster schedule policy is proposed in VMGPURMS to improve the performance of GPU.Evaluation demonstrates that the eciency of GPUs with the GPU cluster schedule policy could be improved by 17.
GPGPU CUDA virtualization virtual machine CPU/GPU hybrid system GPU schedule
国内会议
湖北恩施
英文
1-8
2014-09-13（万方平台首次上网日期，不代表论文的发表时间）