Today, the infrastructure market requires solutions for resource-intensive tasks, that are being flexible enough when it comes to power distribution.
These include GPU passthrough technology implemented in VMmanager virtualization system. It allows connecting graphics accelerators to virtual machines similar to a physical computer.
GPU virtualization (vGPU) has now become available in the platform. It provides distribution of GPU’s computing resources between several users.
The key difference between vGPU and GPU passthrough is that with this technology, you can create virtual copies of the GPU, while GPU passthrough provides direct access to the physical device.
How the technology works:
- The physical GPU is divided into virtual ones (vGPUs), each assigned to a separate virtual machine (VM).
- The hypervisor and vGPU drivers manage the allocation of resources such as memory and computing cores.
- The technology emulates graphics processor for each VM, ensuring shared use of a single device.
Key advantages of vGPU:
- Saving resources – one GPU serves multiple VMs.
- Scalability – suitable for remote desktops (VDI), cloud solutions, and data center rendering.
- Adaptability – simple control and dynamic power distribution.
- Integration to the platform – VM template support, automatic resource deallocation.
Who would benefit from the new product:
- Hosting providers – to provide clients with access to the GPU without allocating separate graphics cards.
- Corporate clients – to implement VDI, test drivers, and work with graphical applications.
- Developers and engineers – for testing and developing software and cloud services, games, solving machine learning problems, rendering, and other scenarios where full GPU power is not required.
vGPU technology is an effective way to optimize IT infrastructure, especially in demand with environments where the rational use of expensive equipment is important.
Please note that vGPU is supported only for some Nvidia GPU devices. For the list of supported devices, see Nvidia documentation.
GPU passthrough and the following platform features are not supported to work together:
- enabling high availability in the cluster if a vGPU VM is on the list of VMs to recover;
- enabling the balancer in the cluster;
- vGPU VM evacuation in cluster node maintenance mode;
- VM snapshot creation;
- VM migration.
In the near future, the VMmanager team is planning to implement vGPU use for cluster nodes running the Astra Linux OS.
For more information about the technology and its configuration, see the ISPsystem product documentation.
For more information about the technology and its configuration, read ISPsystem product documentation.
