It’s the start of a new year, and many in the industry have declared 2014 to be “the year of VDI.” Remember when they said that about 2013? And 2012, 2011, and 2010? Despite proclamations from analysts and industry experts alike, VDI just hasn’t caught on in a big way yet with customers. Sure, there have been advancements that have made VDI easier for IT departments to deploy and manage. But when it comes to the success and adoption of VDI, it is user experience that matters most. Users demand performance, and VDI performance has been limited by insufficient IOPS in storage and graphics. But that’s all about to change.
Tackling the IOPS problem
When it comes to supporting hundreds, if not thousands, of virtual desktops, the more IOPS you have the better. The enterprise-level storage needed to facilitate high IOPS on VDI, particularly with flash accelerated storage, can be very expensive. This added cost makes VDI a tough sell if the cost per user to deploy VDI is two- to four-times more than buying a PC. In late 2012, Atlantis Computing released a product call Atlantis ILIO Diskless VDI to combat this issue. This product eliminated the need for storage by turning server RAM into usable storage for VDI, providing what they claimed to be 300+ IOPS per user. Not only was this much faster than the typical PC user experience, it was also much less expensive than a new storage system. Atlantis had solved two big problems for high-performance VDI: cost and IOPS. But without PC class graphics capabilities, VDI was still underpowered for most high performance workloads. That’s where NVIDIA came in.
NVIDIA Grid K1/K2: The missing piece for high performance VDI
Unlike PCs, virtual desktops do not have the dedicated hardware to process graphics, so rich graphics, video conferencing, and 3D rendering are pretty much impossible. The inability to deliver hardware-accelerated graphics drove many high-performance users away from VDI, especially those with CAD software workloads. For decades, NVIDIA has been a major player in the visual computing space, providing graphics processors for gamers and professionals alike. In Q3 of last year, NVIDIA introduced its GRID K1/K2 graphics boards. GRID boards enable virtualization of graphics processing units (GPUs) across multiple desktops—a major breakthrough for high performance VDI.
With GRID, multiple virtual GPUs share a single NVIDIA GRID GPU (similar to the virtualization of CPUs). This allows direct GPU access from the desktop using native NVIDIA drivers without translation by the hypervisor. And with support for the latest version of Direct X and Open GL, GRID has earned top praise from leading high performance customers—even CAD vendors such as Autodesk and AutoCAD. Currently, Citrix is the only company that can fully utilize this technology when deployed with XenServer and XenDesktop 7, but I expect this to become available to VMware in the near future.
2014: The Year of VDI
There’s no doubt that VDI has made huge strides in the last two years, and it’s becoming apparent that VDI is not just a passing phase. With Atlantis solving the cost and performance issues, NVIDIA GRID is the final piece of the VDI puzzle. This is the beginning of a new era for VDI as it achieves the capabilities needed to leave its mark in today’s fast paced, highly competitive business world. So while 2014 may be the year of VDI, it is really just the beginning.
If you have any questions about VDI or this article, please email Leon Ngo at [email protected]m
This article first appeared in the 2014 issue of Elevate, Zumasys’ Technology Solutions Newsletter.