Bright Cluster Management software will be used on UND’s condominium cluster and its general use research cluster to support research from computational chemistry, weather and climate simulation, health analytics, to image analysis and machine learning.
“UND system administrators have been pleased with Bright Cluster Manager, as even those not experienced with HPC clusters are able to get up to speed with these complex environments quickly,” said Aaron Bergstrom, UND’s high performance computing specialist.
“Our admins particularly like the software’s ease of use and the responsive, professional technical support staff,” Bergstrom said. “They have received quick ticket responses from Bright that work well, and they can often resolve multiple issues the same day.”
“Personnel funding for system admin support staff is limited, so our admins’ time is a valuable commodity,” Bergstrom said. “Bright products allow them to get up to speed quickly, giving us the ability to work with innovative computational environments that we would otherwise not have the time to explore,” said Bergstrom.
Future plans include a Bright OpenStack deployment for cluster-as-a-service (CaaS) applications, enabling HPC computer scientists to build and tear down their own clusters and customize them to their specific needs. This includes not only traditional HPC clusters, but environments for Big Data, such as Hadoop.
As a founding institution in the NSF-funded Midwest Big Data Hub, UND – through its Computational Research Center – is exploring how a private cloud environment based on the Bright OpenStack software can support data science projects related to their big data partnerships.
The university also wants to explore long-tail of science needs for those computational problems that are better served by one-off server instances for short periods, providing researchers who value a traditional desktop with a computational environment that better meets their needs and expectations.
“We feel that having a triad of resources – condominium cluster, general-use cluster, and an OpenStack-based cloud resource – will give us the flexibility to meet the growing computational needs of a broader range of researchers on our campus,” said Bergstrom.
“We are very pleased to be working with the University of North Dakota to help further their cutting edge research,” said Bill Wagner, Bright Computing CEO. “Bright’s management solutions are being used by many institutions that want to combine high-performance computing, big data, deep learning, and private cloud environments within a highly dynamic, flexible IT infrastructure.”