Foreword - By Jason Langone, VCDX #54 @langonej



The world of virtual desktop infrastructure (VDI) has evolved greatly in the last three years. Solutions from VMware and Citrix, for example, are now mainstream and supporting production, daily-use desktops. From trading floors, to call centers, to military personnel, virtual desktops (vDesktops) are becoming an integral component of many organizations.



Historically, vDesktops supported use cases that did not include a heavy video component. This is because the underlying specialized compute resources needed to deliver a positive end user experience were not available for VDI solutions.



There were two reasons for their unavailability: hardware availability and GPU abstraction. Hardware availability refers to there being a lack of hardware (a PCI card with GPU) designed to support the user density of a VDI environment that is also compatible with a popular hypervisor. Nvidia has since emerged as the leader in this space due, in part, to their close partnership with the two major VDI providers, VMware and Citrix. The second reason GPU has been missing in VDI solutions is that for GPU in a vDesktop to make sense the GPUs [likely] need to be shared. This means multiple end users all sharing the processing capability of a given GPU living inside the server host.



Both of these hurdles have been overcome. While GPU in VDI is still in its infancy, there now exists solutions worth discussing, analyzing, testing, and in some cases implementing.



This guide attempts to explain how to determine if the addition of a GPU solution will meet your VDI requirements, and if so, which GPU for VDI configuration is most ideal given your environment.