[Hardware] NVIDIA cards for supercharged virtual desktops

The use of virtual desktops intended for productivity tool users as well as those working remotely is now a common occurrence. These technologies are evolving rapidly so that their uses can be extended, particularly in professions where intensive applications made migration a delicate process until now.

The recent technological advances which enable professionals to take advantage of high performance virtual desktops are linked to the release of graphics cards designed for this purpose, like the NVIDIA K2, which is available at OVH and compatible with VMware and Citrix virtualisation solutions.

Here are some example use cases and their relative applications:

Design, mainly for using software suites for editing and manipulating 3D elements such as SolidWorks, Autodesk Maya and 3ds Max types.

Powerful office automation for users of office suites - including PowerPoint, as well as professionals who need to review the work of designers, for example in SolidWorks View.

Creative/Studio for those needing resources for photo/video centric content applications like Adobe Premiere Pro or Photoshop.

The K1 and K2, Kepler generation, provide allocated GPUs used in the form of vGPUs by virtual desktop solutions (see the box below). The K2 cards are already available on our Big HG servers, allowing customers to build virtual desktop infrastructures as well as work stations with the power to run resource intensive applications.

pGPU & vGPU

These acronyms are used to distinguish between the “physical” part of the GPU – dedicated graphics processor – called the pGPU, several of which can be on the same graphics card (2 in the case of a K2 card) and its "virtual" counterpart, the vGPU. The "virtual" part of the GPU – which is also part of a pGPU – is exploited through a virtualisation solution and allocated to a virtual desktop based on the level of performance required.

The use of GPU cards permits many professions which were previously limited by applications requiring superior performance, to accelerate the adoption of virtual desktops.

Taking things even further, NVIDIA announced last summer at VMworld San Francisco, GRID 2.0, its latest solution dedicated to virtual desktops, which comes in two variety of cards, the M6 and the M60, and this time with Maxwell GPU architecture.

According to NVIDIA, this new generation allows for a greater density of users per card.
Notably, one of the strong points of version 2.0 lies in its compatibility with Linux, which extends the scope of applications eligible for this technology.

These cards have also been developed in collaboration with VMware and Citrix to maximise integration within their suites - Horizon 6 and XenDesktop, respectively - and especially for pass-through, a technology which makes it possible for a virtual desktop infrastructure to harness the complete power of a physical GPU to render 2D or 3D graphics.

Passthrough

This feature allows a virtual machine or a virtual desktop to exploit a physical resource - in this case a graphic processor, without having to go through the hypervisor. Performance is thereby enhanced due to the reduction of intermediaries exchanging data. This technology is particularly used to access raw disks, PCI components, and USB devices.

Citrix and VMware have already integrated these cards into their solutions and have talked about their implementation and benefits: VMware Horizon and XenServer et XenDesktop.