公开(公告)号：US10198299B1

公开(公告)日：2019-02-05

申请号：US15798365

申请日：2017-10-30

Applicant: VMware, Inc.

Inventor： Xin Xu ,  Bryan Tan ,  Wei Xu ,  Tao Ren ,  Radu Rugina ,  Vivek Mohan Thampi

IPC: G06F9/50 ,  G06F13/28 ,  G06F13/16 ,  G06F13/42 ,  G06F9/455

Abstract: Techniques for enabling live migration of VMs with passthrough PCI devices are provided. In one set of embodiments, a hypervisor of a host system can create a copy of a DMA buffer used by a VM of the host system and a passthrough PCI device of the VM. The hypervisor can further designate one of the DMA buffer or the copy of the DMA buffer as a vCPU buffer that is accessible by the VM, and designate the other of the DMA buffer or the copy of the DMA buffer as a device buffer that is accessible by the passthrough PCI device. The hypervisor can then synchronize the vCPU buffer and the device buffer with each other as the VM and passthrough PCI device interact with their respective buffers, and as part of the synchronization can intercept DMA work requests submitted by the VM/completed by the passthrough PCI device.

公开(公告)号：US20190146853A1

公开(公告)日：2019-05-16

申请号：US16223071

申请日：2018-12-17

Applicant: VMware, Inc.

Inventor： Xin Xu ,  Bryan Tan ,  Wei Xu ,  Tao Ren ,  Radu Rugina ,  Vivek Mohan Thampi

IPC: G06F9/50 ,  G06F13/16 ,  G06F13/42 ,  G06F9/455 ,  G06F13/28

Abstract: Techniques for enabling live migration of VMs with passthrough PCI devices are provided. In one set of embodiments, a hypervisor of a host system can create a copy of a DMA buffer used by a VM of the host system and a passthrough PCI device of the VM. The hypervisor can further designate one of the DMA buffer or the copy of the DMA buffer as a vCPU buffer that is accessible by the VM, and designate the other of the DMA buffer or the copy of the DMA buffer as a device buffer that is accessible by the passthrough PCI device. The hypervisor can then synchronize the vCPU buffer and the device buffer with each other as the VM and passthrough PCI device interact with their respective buffers, and as part of the synchronization can intercept DMA work requests submitted by the VM/completed by the passthrough PCI device.

公开(公告)号：US11681544B2

公开(公告)日：2023-06-20

申请号：US17395147

申请日：2021-08-05

Applicant: VMware, Inc.

Inventor： Xin Xu ,  Na Zhang ,  Xiaolong Cui ,  Jiayuan He ,  Ridhi Surana

IPC: G06F9/455 ,  G06N20/00 ,  G06F11/34 ,  G06F11/30

CPC classification number: G06F9/45558 ,  G06F11/301 ,  G06F11/3433 ,  G06N20/00 ,  G06F2009/45595

Abstract: Disclosed are aspects of interference-aware virtual machine assignment for systems that include graphics processing units (GPUs) that are virtual GPU (vGPU) enabled. In some examples, an interference function is used to predict interference for assignment of a workload to a graphics processing unit (GPU). The interference function outputs a predicted interference to place the workload on the GPU. The workload is assigned to the GPU based on a comparison of the predicted interference to a plurality of predicted interferences for the workload on various GPUs.

公开(公告)号：US20210373930A1

公开(公告)日：2021-12-02

申请号：US17395147

申请日：2021-08-05

Applicant: VMware, Inc.

Inventor： Xin Xu ,  Na Zhang ,  Xiaolong Cui ,  Jiayuan He ,  Ridhi Surana

IPC: G06F9/455 ,  G06N20/00 ,  G06F11/34 ,  G06F11/30

Abstract: Disclosed are aspects of interference-aware virtual machine assignment for systems that include graphics processing units (GPUs) that are virtual GPU (vGPU) enabled. In some examples, an interference function is used to predict interference for assignment of a workload to a graphics processing unit (GPU). The interference function outputs a predicted interference to place the workload on the GPU. The workload is assigned to the GPU based on a comparison of the predicted interference to a plurality of predicted interferences for the workload on various GPUs.

公开(公告)号：US11113093B2

公开(公告)日：2021-09-07

申请号：US16432108

申请日：2019-06-05

Applicant: VMware, Inc.

Inventor： Xin Xu ,  Na Zhang ,  Xiaolong Cui ,  Jiayuan He ,  Ridhi Surana

IPC: G06F9/455 ,  G06N20/00 ,  G06F11/34 ,  G06F11/30

Abstract: Disclosed are aspects of interference-aware virtual machine assignment for systems that include graphics processing units (GPUs) that are virtual GPU (vGPU) enabled. In some examples, a plurality of workloads are executed alone and co-located with other workloads in a virtual graphics processing unit (vGPU)-enabled system to determine baseline parameters and measured interferences. A machine learning model is trained to predict interference based on the measured interferences and the baseline parameters. A workload is assigned and executed on a particular GPU associated with a minimum predicted interference with the workload based on currently-assigned workloads of the particular GPU.

公开(公告)号：US09898430B2

公开(公告)日：2018-02-20

申请号：US14539528

申请日：2014-11-12

Applicant: VMware, Inc.

Inventor： Bhavesh Davda ,  Xin Xu ,  Guolin Yang

IPC: G06F13/28 ,  G06F13/24 ,  G06F9/455

CPC classification number: G06F13/28 ,  G06F9/45558 ,  G06F13/24 ,  G06F2009/45579 ,  G06F2009/45583

Abstract: Techniques for tracking, by a host system, virtual machine (VM) memory modified by a physical input/output (I/O) device that supports I/O virtualization are provided. In one embodiment, a hypervisor of the host system can receive a hardware interrupt from the physical I/O device, where the hardware interrupt indicates that a virtual function (VF) of the physical I/O device has completed a direct memory access (DMA) write to a guest memory space of a VM running on the host system. In response to the hardware interrupt, the hypervisor can invoke a function implemented by a physical function (PF) driver of the physical I/O device, where the function is configured to inspect the VF's state in order to identify memory portions modified by the DMA write. The hypervisor can then mark, in a hypervisor-level page table, one or more memory pages corresponding to the identified memory portions as dirty pages.

公开(公告)号：US20160132443A1

公开(公告)日：2016-05-12

申请号：US14539528

申请日：2014-11-12

Applicant: VMware, Inc.

Inventor： Bhavesh Davda ,  Xin Xu ,  Guolin Yang

IPC: G06F13/28 ,  G06F13/24 ,  G06F9/455

CPC classification number: G06F13/28 ,  G06F9/45558 ,  G06F13/24 ,  G06F2009/45579 ,  G06F2009/45583

Abstract: Techniques for tracking, by a host system, virtual machine (VM) memory modified by a physical input/output (I/O) device that supports I/O virtualization are provided. In one embodiment, a hypervisor of the host system can receive a hardware interrupt from the physical I/O device, where the hardware interrupt indicates that a virtual function (VF) of the physical I/O device has completed a direct memory access (DMA) write to a guest memory space of a VM running on the host system. In response to the hardware interrupt, the hypervisor can invoke a function implemented by a physical function (PF) driver of the physical I/O device, where the function is configured to inspect the VF's state in order to identify memory portions modified by the DMA write. The hypervisor can then mark, in a hypervisor-level page table, one or more memory pages corresponding to the identified memory portions as dirty pages.

Abstract translation: 提供了由主机系统跟踪由支持I / O虚拟化的物理输入/输出（I / O）设备修改的虚拟机（VM）存储器的技术。 在一个实施例中，主机系统的管理程序可以从物理I / O设备接收硬件中断，其中硬件中断指示物理I / O设备的虚拟功能（VF）已经完成了直接存储器访问（DMA ）写入在主机系统上运行的VM的guest虚拟机内存空间。 响应于硬件中断，虚拟机管理程序可以调用由物理I / O设备的物理功能（PF）驱动器实现的功能，其中功能被配置为检查VF的状态，以便识别由DMA修改的存储器部分 写。 管理程序可以在管理程序级页面表中将与所标识的存储器部分相对应的一个或多个存储器页面标记为脏页。