Create a Custom VM Class for NVIDIA vGPU Devices

Refer to this topic to create a custom VM class for NVIDIA Grid vGPU devices.

Create a Custom VM Class with vGPU Profile (v8 U2 P03 and later)

NVIDIA vGPU (Virtual Graphics Processing Unit) allows multiple virtual machines (VMs) to share a single physical GPU. To use vGPUs with TKGS clusters, you define a custom VM class. Starting with this release there is a new wizard for defining custom VM classes. Unlike the previous method for defining a custom VM class, the vGPU profile is read from the device instead of being configured in the VM class.

The VM Operator polls the vCenter inventory to obtain all of the vGPU devices installed on ESXi hosts comprising the vSphere Cluster where Supervisor is deployed. The vGPU device defines its profile. The vGPU device name indicates if the profile is multi-instance GPU (MIG) or time shared GPU. MIG slices the compute and enables multiple workloads to run in parallel on a single GPU. Time sharing provides shared access to the GPU. MIG mode is based on a newer GPU architecture and is only supported on NVIDIA A100 and A30 devices. Refer to the NVIDIA documentation for more information.

For example, the GPU device "grid-a100-40c" provides a time shared vGPU profile that allocates an NVIDIA A100 GPU device with 40 GB of memory to a VM. The equivalent MIG-based vGPU profile would be the "grid-a100-7-40c" device. You can identify that it is a MIG profile because there is an extra number between the device and the RAM. The "7" indicates there are 7 compute slices on the GPU device. MIG-based vGPU profiles can have 1, 2, 3 or 7 compute slices.

1. From the vSphere Client home menu, select Workload Management > Services.
2. Select the VM Classes tab.
3. Click Create VM Class.

   This action launches the Create VM Class wizard that walks you through creating a VM class.

4. For Name, enter a name for the VM class and click Next.
   The VM class name identifies the VM class. Enter a unique DNS compliant name that follows these requirements:

   • Use a unique name that does not duplicate the names of default or custom VM classes in your environment.
   • Use alphanumeric string with maximum length of 63 characters.
   • Do not use uppercase letters or spaces.
   • Use a dash anywhere except as a first or last character. For example, vm-class1.
   • After you create the VM class, you cannot change its name.
5. For Compatibility, select ESXi 8.0 U2 and later and click Next.
   For more information, see Virtual Machine Compatibility.

   Note: You cannot change the hardware compatibility of a VM class after it is created.
6. For Configuration > Virtual Hardware, add the NVIDIA GPU device to the VM class.
   1. Select Configuration > Virtual Hardware > Add New Device > PCI Device.
   2. Select the desired NVIDIA Grid vGPU device from the list. There are two types of NVIDIA Grid vGPU profiles: Time Sharing and Multi-Instance GPU Sharing. The profile is detected by the system when you select the device.
      Note: You can add only one NVIDIA GRID vGPU device of type MIG profile to a VM class.
   3. Click Select and the New PCI device appears in the Virtual Hardware tab.
7. For Configuration > Virtual Hardware, specify the desired settings for CPU, Memory, New PCI Device, Video Card, and Security Devices.

   Table 1. CPU Configuration

   Setting	Configuration
   CPU	Select the number of virtual CPUs for the VM. See Virtual CPU Configuration and Limitations for more information.
   CPU Topology	Assigned at power on
   Reservation	Reservation must be between 0 and 10 MHz
   Limit	Limit must be greater than or equal to 10 MHz
   Shares	Options are Low, Normal, High, Custom
   Hardware virtualization	Select this option to expose hardware assisted virtualization to the guest OS
   Performance Counters	Enable virtualized CPU performance counters
   Scheduling Affinity	Select a physical processor affinity for this virtual machine. Use '-' for ranges and ',' to separate values. For example, "0, 2, 4-7" would indicate processors 0, 2, 4, 5, 6 and 7. Clear the string to remove affinity settings.
   I/O MMU	Select to enable memory management unit (page to disk)

   Table 2. Memory Configuration

   Setting	Configuration
   Memory	Select the size of the memory for the VM. See Maximum Virtual Machine Memory for more information.
   Reservation	Specify the guaranteed minimum allocation for a virtual machine, or reserve all guest memory. If the reservation cannot be met, the VM cannot run.
   Limit	Select the amount of memory to limit to place a limit on the consumption of memory for a VM.
   Shares	Select the amount of memory to share. Shares represent a relative metric for allocating memory capacity. For more information, see Memory Sharing.
   Memory Hot Plug	Enable (check) to allow the addition of memory resources to a VM that is powered on. See Memory Hot Add Settings for details.

   Table 3. New PCI Device > GPU Sharing Configuration

   Time Sharing Mode	MIG Mode
   In Time Sharing mode, the vGPU scheduler instructs the GPU to perform the work for each vGPU-enabled VM serially for a duration of time with the best effort goal of balancing performance across vGPUs.	MIG mode allows multiple vGPU-enabled VMs to run in parallel on a single GPU device. If you do not see the MIG option, the PCI device you selected does not support it.

   Table 4. Configure Video Card

   Setting	Configuration
   Video Card	Choose to auto-detect settings from the hardware or enter custom settings. If you select auto-detect, other settings are not configurable.
   Number of displays	Select the number of displays.
   Total video memory	Enter the total video memory, in MB.
   3D Graphics	Select to enable 3D support.

   Table 5. Configure Security Devices

   Settings	Configuration
   Security Device	If the SGX security device is installed, you can configure the VM settings here, otherwise this field is not configurable. See the SGX documentation for details.

8. Select the Configuration > VM Options tab and configure any additional VM settings. See Configuring Virtual Machine Options for guidance.
9. Select the Configuration > Advanced Parameters tab and add any attributes for the VM class.
10. Click Next.
11. At the Review and Confirm page, review the details and click Finish.
12. Associate the new VM class with the vSphere Namespace. See Associate VM Classes with the vSphere Namespace.

Figure 1. NVIDIA vGPU Device Selection

Figure 2. NVIDIA vGPU New PCI Device

Create a Custom VM Class with a vGPU Profile (v8 U2 and earlier)

To next step is to create a custom VM Class with a vGPU profile. The system will use this class definition when it creates the TKG cluster nodes.

Follow the instructions below to create a custom VM Class with a vGPU profile.

1. Log in to the vCenter Server using the vSphere Client.
2. Select Workload Management.
3. Select Services.
4. Select VM Classes.
5. Click Create VM Class.
6. At the Configuration tab, configure the custom VM Class.

   Configuration Field	Description
   Name	Enter a self-descriptive name for the custom VM class, such as vmclass-vgpu-1.
   vCPU Count	2
   CPU Resource Reservation	Optional, OK to leave blank
   Memory	80 GB, for example
   Memory Resource Reservation	100% (mandatory when PCI devices are configured in a VM Class)
   PCI Devices	Yes Note: Selecting Yes for PCI Devices tells the system you are using a GPU device and changes the VM Class configuration to support vGPU configuration. For more information, see Add PCI Devices to a VM Class in vSphere with Tanzu.

   For example:

   

7. Click Next.
8. At the PCI Devices tab, select the Add PCI Device > NVIDIA vGPU option.
9. Configure the NVIDIA vGPU model.

   NVIDIA vGPU Field	Description
   Model	Select the NVIDIA GPU hardware device model from those available in the NVIDIA vGPU > Model menu. If the system does not show any profiles, none of the hosts in the cluster have supported PCI devices.
   GPU Sharing	This setting defines how the GPU device is shared across GPU-enabled VMs. There are two types of vGPU implementations: Time Sharing and Multi-Instance GPU Sharing. In Time Sharing mode, the vGPU scheduler instructs the GPU to perform the work for each vGPU-enabled VM serially for a duration of time with the best effort goal of balancing performance across vGPUs. MIG mode allows multiple vGPU-enabled VMs to run in parallel on a single GPU device. MIG mode is based on a newer GPU architecture and is only supported on NVIDIA A100 and A30 devices. If you do not see the MIG option, the PCI device you selected does not support it.
   GPU Mode	Compute
   GPU Memory	8 GB, for example
   Number of vGPUs	1, for example

   For example, here is a NVIDIA vGPU profile configured in Time Sharing mode:

   

   For example, here is a NVIDIA vGPU profile configured in MIG mode with supported GPU device:

   

10. Click Next.
11. Review and confirm your selections.
12. Click Finish.
13. Verify that the new custom VM Class is available in the list of VM Classes.

vGPU with Dynamic DirectPath IO

If you are using vGPU with Dynamic DirectPath IO, complete the following additional configuration. Add a second PCI Device configuration to the custom VM class you created with Dynamic DirectPath IO specified and the supported PCI Device selected. When a VM Class of this type is instantiated, the vSphere Distributed Resource Scheduler (DRS) determines VM placement.

1. Select Workload Management.
2. Select Services.
3. Select VM Classes.
4. Edit the custom VM Class that is already configured with the NVIDIA vGPU profile.
5. Select the PCI Devices tab.
6. Click Add PCI Device.
7. Select the Dynamic DirectPath IO option.

   

8. Select the PCI Device.

   For example:

   
9. Click Next.
10. Review and confirm your selections.
11. Click Finish.
12. Verify that the new custom VM Class is available in the list of VM Classes.