GRID Community Perspective – Part II: How NVIDIA Empowers Education Through GPU Virtualization

NGCA Member Tobias Kreidl shares his experiences implementing NVIDIA GRID at Northern Arizona University
by Tony Paikeday

By Tobias Kreidl

Note: This is part two of a guest series from Tobias Kriedl, an NVIDIA GRID Community Advisor (NGCA) and Academic Computing Team Lead at Northern Arizona University.  Tobias brings considerable experience in using a range of NVIDIA technologies for HPC, VDI, and graphics as a long time user of Quadro, Tesla and GRID products.

In my last post I wrote about Northern Arizona University (NAU) and our experience with NVIDIA GRID. Numerous educational institutions have already embraced GRID technology. A dozen or more successful case studies have been published. With GRID, NVIDIA’s new licensing model supports licensing of the M-series GPUs. Because typical student end users require smaller configurations, costs can be saved by making use of more vGPUs per board or via GPU passthrough, plus with the added power and memory buffers of the M10 and M60 units, fewer of those will be needed compared to their K1 and K2 predecessors. Furthermore, with new monitoring tools available (see, for example, this blog regarding VGPU monitoring), usage can be carefully watched, sizing adjusted and future plans made based on these metrics.

The emphasis at many educational institutions is on applications that profit the most from GPU back ends and clients may include classrooms, labs and BYOD devices, as well as possible offices or other locations that contain lower-end workstations. There is much awareness of how GPUs can benefit just about any imaginable application, which is certainly an interesting option for those who wish to focus on low-end workstations. It’s also imperative that the workforce be properly trained to embrace these new technologies, so providing the means for students to become proficient in both using them and in some cases, learning how to configure and administer them, will ensure better integration of such individuals into the corporate world upon completion of their studies.

One very early adopter – also involved in much of the beta testing of GRID – was the College of Engineering at the Georgia Institute of Technology. Cost savings and the convenience of hosting GPU-intensive calculations on servers and thereby reducing the requirements for expensive desktop hardware were immediate benefits. Another was the ability to access the same computational and graphical display experience from virtually any device using remote access protocols. Students, faculty and staff were no longer shackled to limited access times to labs nor to any physical location. They could work out of homes, dorms, cafes or with wireless connectivity, even from the outdoors.

There are a number of educational institutions that exemplify what GRID technology has been able to accomplish. To illustrate the diversity in the directions educational institutions are moving towards, I have solicited input from a few selected institutions.

Didier Contis, IT Director, Georgia Institute of Technology, College of Engineering (https://coe.gatech.edu) was one of the earliest adopters of GRID technology.  With several dozen GRID K1 and GRID K2 GPUs, they support over 13,000 engineering students. Details about Georgia Tech’s Engineering schools and the evolution of their computing facilities can be gleaned from a number of articles, including the following where the development and expansion over the years are clearly demonstrated:

Right now, they are looking at continuing the buildup of facilities by mixing Tesla  M10 and M60 GPUs, where the M10 will cover around 80% of their needs and the M60 units will target high-end use cases. Currently, the M10/M60 mix is hosted on Dell R730 servers. Mr. Contis states, “The biggest challenge is to determine the correct ratio of Virtual PC/Workstation/Vapps licenses.” The department is currently running tests with 2 x M60 and 1 x M10.

René Bigler, Head of IT at the Berufsbildungszentrum IDM – Industrie Dienstleistung Modegestaltung, (https://idm.ch) in Switzerland foresees that graphics-accelerated virtual desktops in education will provide a boost for the user’s virtual workspace. He writes, “With the release of Windows 10 and Server 2016, we are planning to upgrade our existing Citrix XenDesktop-based VDI environment. At the same time, we are testing how GPUs would improve the user experience since more and more applications and the new operating systems themselves can benefit from GPU acceleration.”

“Our students and faculty access the virtual desktops and a wide variety of published apps, including some CAD applications and tools for image editing from different types of endpoints. There are Linux-driven thin clients, private and corporate Windows and Mac clients, as well as many ChromeBooks and different mobile devices such as tablets and smartphones.”

“For our proof of concept we opted for the new NVIDIA Tesla M10 in conjunction with the GRID Virtual PC licenses. For the hypervisor we will use XenServer 7 (included with our XenDesktop licensing) on a Dell PowerEdge R730xd with two 10-core Xeon CPUs and 384 GB RAM. This high-density setup will allow us to deliver up to 64 vGPU accelerated virtual desktops to our users.”

“In education there is more than just the basic task worker workloads such as Office. Students and faculty users have an increasing demand for multimedia content and heavier graphic applications in their daily digital learning requirements—this is where our VDI setup has to keep pace and deliver the best user experience possible. Our goal is that the user will see no difference between his virtual and physical workspace.”

Eric Hosmer, System Administrator, Sr., at Kettering University (https://kettering.edu) sums up the position of his university as they embark on adopting GRID 2.0 technology, “Kettering University is national leader in experiential STEM (science, technology, engineering and math) and business education. This creates a need for IT to support a very diverse mix of graphics-intensive software across all departments.”

“Additionally, because all Kettering University students work in co-op terms with corporate partners during alternating terms, there is increased pressure to support software used by their employers. These experiences have shown the University’s IT department that graphics matter with all software. Bring Your Own Device (BYOD) generates new challenges and the need to support graphic-intensive software inspires innovation.”

“In 2014, we launched Citrix XenDesktop, the solution for BYOD and users’ mobility, and our users loved it. At that time we had to leave out software that needed more graphics. The hardware to support Citrix HDX 3D Pro had too high a cost with too low of a user density. With NVIDIA Grid came the NVIDIA M60, supporting up to 32 virtual desktops each with its own vGPU; the M10 doubles the number to 64. Kettering plans on adding M10s to all of our hosting servers (CPU 2 x Intel Xeon E5-2697A v4, 2.6 GHz (16-Cores) with 512GB of DDR4-2400 RAM). This will give all of our virtual desktops vGPUs. Kettering University IT can then deliver the user experience our students, faculty and staff have come to expect in our physical labs. More importantly, IT can provide our users all of those graphic-intensive software components that were missing from our virtual labs prior to GRID cards.”

“Tethering users to physical PC labs across campus does not foster the best collaboration or creativity. Giving users those missing software and better user experiences via GRID provides user mobility, producing more synergy for collaboration and creativity.”

As you can see, NVIDIA GRID is bound to play an increasing role in education and as seen clearly from some of the remarks from institutions who have gone this route, are in the middle of deployment or are planning to make use of GPU technologies, there is a lot of interest and enthusiasm about deploying GRID technology. It is a pleasure and privilege to help support such efforts and I am excited about those as well as our own prospects at NAU.

 

Tobias Kreidl

TOBIAS KREIDL, PhD

ACADEMIC COMPUTING TEAM LEAD, NORTHERN ARIZONA UNIVERSITY

Tobias spent 15 years as a research astronomer before a career charge that took him to NAU’s (Northern Arizona University) central IT department. A strong advocate for the use of computers in education, he has helped introduce a number of technologies to students and staff, embracing many open source tools, as well as Citrix and other commercial products. An early advocate of virtual computing, he is a leading poster on the Citrix discussion forum and works with his team on everything from virtual servers and thin clients to software defined storage, GPU technologies, disaster recovery, Web-based services, and anything else that can enhance and secure the educational computing environment. Tobias brings considerable experience in using a range of NVIDIA technologies for HPC, VDI, and graphics as a long time user of Quadro, Tesla and GRID products.

To learn more about the NGCA and how you can become an advisor, click here.