Nor-Tech Leadership Innovates Scheduling Solution to Fully Leverage GPU Resources

Share Article

Nor-Tech engineers just innovated a solution for seamlessly scheduling GPU resources—dramatically reducing operational inefficiencies and saving significant development time.

Nor-Tech President and CEO David Bollig

Nor-Tech President and CEO David Bollig

Our star engineering team is among the advantages that separates us from the competition.

Nor-Tech engineers just innovated a revolutionary solution for seamlessly scheduling GPU resources—dramatically reducing operational inefficiencies and saving significant development time. Before this, most organizations had to create their own proprietary custom code, which was very expensive to develop.

The engineering team, led by Vice President of Engineering Dom Daninger, was faced with a pervasive issue: how to eliminate the scheduling issues associated with large numbers of GPU accelerators. In this case, it was a client with simulation applications.

Daninger explained, “The client’s code developers were working on moving some of the simulation and modeling code over to GPU acceleration from what had previously been all CPU executed code. They had six servers each with eight NVIDIA GPUs. None of the GPUs had any video display attached; they were purely computational. Up until then, this client was using a resource manager that was not able to manage multiple GPU resources. So the resource manager had no way to determine which GPU resources were available and which were busy.”

Nor-Tech recommended PBS PRO Open, which is capable of managing two forms of accelerator devices, GPUs and Intel Xeon-Phi. The Nor-Tech team spent many hours developing a highly custom installation of PBS Pro Open to steer individual GPU jobs to individual GPUs within each server, not just for this client but for Nor-Tech clients in similar situations down the road.

PBS Pro Open (which also comes in a commercial version) has two node resources that it can manage: natural and V-node. Nor-Tech’s solution included eight virtual nodes within each one of the client’s six servers. Each virtual node was comprised of a small amount of CPU resource and one of the physical GPUs. The software used to run those GPUs was NVIDIA’s CUDA. Nor-Tech coordinated the individual components to allow one of the CUDA devices to start a job, then use PBS Pro Open to find the job and steer the subsequent incoming jobs to other GPU/CPU resources. Basically they were able to add critical scheduling functionality with a prologue script to PBS Pro Open. “This allows enterprises to fully leverage their GPU resources by enabling them to run in an embarrassing parallel fashion,” Daninger said. “It solves major performance issues and is a big timesaver.”

Nor-Tech prides itself on being the source of many innovative high performance computing innovations such as an HPC cluster chassis reconfiguration and system redesign to accommodate consumer grade GPUs—which reduces the cost appreciably without significantly affecting performance in most applications. Another innovation is the development of a demo cluster to show the ROI of upgrading from a workstation to a cluster for simulation and modeling applications.

Nor-Tech President and CEO David Bollig said, “Our star engineering team is among the advantages that separates us from the competition. In this case, instead of leaving the client to develop their own code, which would have involved a time delay and significant expense on their part, we leveraged our brain trust to come up with a solution that would solve the problem once and for all.”

For a limited time, Nor-Tech will send a $25 Amazon gift card to those interested in learning more about the demo cluster; complete the contact form at http://www.nor-tech.com/solutions/hpc/demo-cluster.

    2016 HPCwire award finalist, Nor-Tech is renowned throughout the scientific, academic, and business communities for easy to deploy turnkey clusters and expert, no wait time support. All of Nor-Tech’s technology is made by Nor-Tech in Minnesota and supported by Nor-Tech around the world. In addition to HPC clusters, Nor-Tech’s custom technology includes workstations, desktops, and servers for a range of applications including CAE, CFD, and FEA. Nor-Tech engineers average 20+ years of experience and are responsible for significant high performance computing innovations. The company has been in business since 1998 and is headquartered in Burnsville, Minn. just outside of Minneapolis. To contact Nor-Tech call 952-808-1000/toll free: 877-808-1010 or visit http://www.nor-tech.com. Full release at: http://www.nor-tech.com/category/news/

Share article on social media or email:

View article via:

Pdf Print

Contact Author

Jeanna Rensselar
@NorTechHPC
since: 12/2010
Follow >
Nor-Tech

Visit website

Media

Nor-Tech logo