Introduction to VMware GPU Virtualization

GPU Virtualization is targeted at a number of research and development and design staff on desktop virtualization that require large 3D designs that do not meet their primary needs with ordinary desktop virtualization. Therefore, it is necessary to increase the GPU on the virtualization platform by means of GPU virtualization.

VMware's GPU virtualization is currently divided into the following three ways:

VSGA:

Refers to "virtual shared graphics acceleration," which is an accelerated VM. The function of VSGA is to provide hardware-accelerated 3D graphics for many virtual desktops using a native-mounted physical GPU (graphics processing unit, also known as a graphics card) in an ESXi host. By providing this capability, you can further extend the usage scenarios and target users that are available with virtual desktops. In addition to extending the target usage scenario, you can further enhance the user experience by providing a more vivid visual experience. VSGA uses VMware drivers in virtual machines, with up to 512MB per user memory.

VGPU:

Refers to the virtual Graphics processing Unit (VGPU), which is a VMware The new feature on vSphere6.0, which is between VSGA and Vdga, is that it can use the core resources of the GPU and can be shared with different configurations, thus achieving a good balance between economy and performance, using the native driver of the GPU in the virtual machine, which is more robust than the VSGA, and through sharing on the video memory to Enables a GPU card to be shared with more users for a more economical purpose. VGPU has different types of configuration files, including the k120q,k220q,k140q,k240q,k160q,k260q,k180q,k280q model, and the application has added a list of certifications to them.

vDGA:

Refers to virtual dedicated graphics acceleration (VDGA), a proprietary hardware graphics acceleration feature provided by VMware Horizon View, used for user scenarios that provide high-end workstation graphics for use in a standalone GPU. When using VDGA, use VMware DIRECTPATHI/O to assign the graphics adapter installed in the underlying host to the virtual machine. Assign a standalone GPU resource to a virtual machine so that the entire GPU is dedicated to that virtual machine. The VDGA is expensive to implement, but can reduce costs compared to a single high-end workstation. The number of virtual machines per host is limited by the number of PCIe x16 slots in that server. There are now server hardware with up to 4 PCIe x16 slots, and the rack also has space for high-end GPUs. Some blade server chassis hardware vendors also offer "side-plug" expansion units capable of supporting up to 8 GPUs. When using VDGA, the driver software from the GPU vendor will be used in the virtual machine.

Three ways to use the scene such as:

650) this.width=650; "title=" Vgpu1.png "alt=" Wkiom1abqx_y3slgaac1okayjiy010.png "src=" http://s1.51cto.com/wyfs02/ M01/78/b1/wkiom1abqx_y3slgaac1okayjiy010.png "/>

The next section will introduce the main technical parameters of VGPU

This article is from the "Chenzhonghua" blog, make sure to keep this source http://chenzhonghua.blog.51cto.com/128732/1729341

Introduction to VMware GPU Virtualization