Accelerated Science: GPU Cluster Case Study

Print Friendly, PDF & Email

This is the sixth article in a series on six strategies for maximizing GPU clusters. This GPU cluster case study and the best practices can will be helpful as you build plans for cloud based GPU processing.

A successful example of how a well-managed GPU cluster allowed scientist to focus on obtaining results comes from the Tokyo University of Agriculture and Technology (TUAT) results. A research group lead by Dr. Akinori Yamanaka develops computation models and simulates engineering materials, for a variety of applications, using HPC. Using Bright Cluster Manager, Dr. Yamanaka and his team were able to immediately focus on algorithm development and not burden the team with cluster administration issues. To learn more about how Bright Cluster Manager accelerated computational research, consult the full case study.


To fully take advantage of NVIDIA GPUs requires several sound strategies. The goal of any HPC resource should be to increase the productivity of researchers and engineers because minimizing time to solution is the goal of many leading HPC installations. Keeping users and developers focused on applications is one of the way to increase productivity and minimize wasted time.

The following is a summary of the strategies discussed in this series. They provide a set of best practices that eliminate many of the delays and challenges facing GPU clusters.
1. Provide a unified system that allows users and developers to focus on applications and coding. Keep infrastructure and tool chain management issues away from those who need to get work done.
2. Keep software current so new features and capabilities are immediately available to users and developers.
3. Manage the users GPU development options with the Modules package to reduce conflicts and misconfigurations.
4. Provide seamless support for all GPU programming models. Developers like options. Provide a well managed and up-to-date development tool chains for all the popular GPU programming methods.
5. Plan for Big Data and Hadoop usage. Breakthroughs in deep learning on GPUs will spawn new applications in this area.
6. Now that GPUs are available in the cloud, develop an approach that allows your organization to take advantage of this resource.

Bright Cluster Manager addresses the above six strategies and provides a comprehensive environment for production GPU cluster computing. There is no reason why a cluster equipped with NVIDIA GPUs cannot operate as an integrated HPC resource.

You can download the complete insideHPC Guide to Managing GPU Clusters courtesy of NVIDIA and Bright Computing.