A GRID Behind Your Cloud

Cloud servers

“Virtual infrastructure is only a piece of the Cloud puzzle…”


ost people have seen the frequent use of (and sometimes overused) IT buzzword “The Cloud”. Many companies have either gone through, or plan to go through an initiative to reduce capital costs and invest in the operational side of the business by migrating most if not all of their IT infrastructure to the Cloud. What many have come to realize is the Cloud can save companies money while adding resiliency.

cloud1While the Cloud may sound like a Wonderland for a company’s IT infrastructure to move too, it has its challenges that have been hard to design around. Cloud Architects have the challenge of trying to design a Virtual Instance for the end-user that needs the highest powered, most graphically intensive machine available – The Engineers. This Designer or Power User has been tied to high-end physical workstations running only the top-end Graphics Cards (GPUs) in order to support high-resolution 3-D imaging and modeling on multiple displays at once. Cloud Service Providers have struggled to get these Power Users into the cloud.

That is until NVIDIA, an American-based graphics card manufacturer, released its latest Kepler GPU architecture and NVIDIA GRID™ portfolio, which includes a series of Server-class GPUs that can be virtualized and shared across a number of users in both gaming and business markets. In the Enterprise space, the GRID solution can leverage virtualized solutions from Citrix, VMware, and Microsoft. GRID allows even the most graphic intensive end-users the ability to move to the Cloud. To the end-user, they are centralizing their engineering software workspace and are providing more security and protection than they had with a physical workspace. The Engineers’ highly responsive Virtual Desktop can be accessed from anywhere, including a tablet or smartphone.


Huang and Templeton announce NVIDIA GRID vGPU on Citrix XenDesktop 7 at Citrix Synergy

I was introduced to this technology last May at Citrix Synergy in Los Angeles. NVIDIA CEO and Co-founder Jen-Hsun Huang talked with Citrix CEO Mark Templeton about their partnership and challenge of designing a virtual workspace for Boeing to be able design the 787 Dreamliner completely digital with thousands of engineers and suppliers across the world collaborating.

What came out of this partnership were major enhancements in the NVIDIA GPU, as well as new high-definition functionality in the Citrix Desktop software. Needless to say, I was impressed when the demonstrator pulled out his fourth-generation iPad and opened up Adobe Photoshop from a Virtual desktop located at a service provider halfway across the country, and proceeded to manipulate photos and drawings. He then opened his iMac and pulled up five enterprise-class applications (video editing, CAD, Illustrator, etc.) that were published from the cloud, and proceeded to modify data in each of those. It was a very powerful demonstration, but as most technicians tend to be, I was skeptical of it in production. I needed to see it done on a larger scale.

…the future of the cloud is bright and the days of saying, “The cloud is just not for you,” are becoming fewer and farther between.



Fast-forward a year and a half. While the NVIDIA GRID hardware is not yet a mainstream piece of hardware installed on every host or included on each and every VDI or Virtual Server sold, it has been adopted by some of the largest Engineering and Architecture firms, as well as companies that have a need to provide a workstation like Virtual desktop to its end-users. These companies are developing what they call VDI 2.0, and have built their virtual backend on the NVIDIA GRID K1 card (which includes Four GK107 GPUs – total 768 CUDA cores, 16 GB DDR3 onboard RAM).



This card is designed for a higher density of users. The NVIDIA GRID K1 technology with vSGA (sharing the virtualized GPU) allows each NVIDIA GRID K1 card to support up to 40 simultaneous and 90 total users. NVIDIA also offers a higher end GRID K2 Card (which includes Two GK104 GPUs – total 3072 CUDA cores, 8 GB DDR5 onboard RAM). The K2 card was made for the Engineer who creates and works with complicated datasets using graphic-intensive applications, such as 3-D design, medical diagnostics, etc.

It is important to understand that the virtual infrastructure is only a piece of the Cloud puzzle. High-resolution displays will only run properly with adequate network bandwidth. And more network bandwidth plus higher end virtual infrastructure almost always equals a higher Operational Cost. But the future of the Cloud is bright and the days of saying, “The Cloud is just not for you,” are becoming fewer and farther between.

Freedom and flexibility have created a paradigm shift in many of the things we do and products we use. The Cloud brings that full-circle with corporate workspaces and NVIDIA giving engineers another tool to make this all possible.

Chris Massey DRS

Chris Massey

Director of Product Development at DRS LLC