Advanced Computing in the Age of AI | Tuesday, May 21, 2024

Nvidia Acquires Bright Computing for its HPC Cluster Management Software 

GPU-maker Nvidia is broadening its enterprise software portfolio by buying HPC cluster management company Bright Computing for an undisclosed sum.

Unlike Nvidia’s ongoing effort to purchase semiconductor IP company Arm Ltd., which has been stymied by regulatory challenges in the U.S., Europe and the U.K., the Bright deal is a straightforward acquisition that aims to expand new opportunities across Nvidia’s customer base and broaden Bright’s reach, according to Nvidia.

“When we started talking to [Bright] on the acquisition side, our companies’ structure, our values, and the way we do support, the way we do service and sales were identical, the way we go to market, it was very similar too, so it just made a lot of sense,” said Charlie Boyle, VP and GM of DGX systems at Nvidia, in an interview with HPCwire. “It is going to both enable the Bright ecosystem to grow, not only on-premises, but in hybrid cloud, which is a huge expansion area. And it is also going to help on the Nvidia side, on core tools that customers have asked us about, but that we just did not have the expertise to develop in house or did not have the time to do because we were focused on accelerated computing, not necessarily heterogeneous systems. Now, we just inherit all of that.”

Spun out of ClusterVision in 2009, Bright Computing makes software that automates the building and management of Linux clusters, both on-premises, in the cloud and across hybrid infrastructures. Nvidia has been working with Bright for more than a decade, integrating Bright Cluster Manager with Nvidia GPUs, CUDA and most recently the company’s DGX systems. “Now we see an opportunity to combine our system software capabilities to make HPC datacenters easier to buy, build and operate, creating a much larger future for HPC,” wrote Boyle in a company blog post.

Bright Cluster Manager is sold via a subscription and support model by channel and OEM partners. Nvidia said that this business model will continue unchanged. “The same people that could sell Bright before can still sell Bright with no change to the customer,” said Boyle. Expanded access via the Nvidia partner network (NPN) will happen within this quarter, Boyle said, and other synergies such as bundling and co-support are being explored.

Nvidia declined to give a headcount for Bright’s workforce but said that all Bright employees will remain and transition to Nvidia.

Charlie Boyle of Nvidia

“Part of the reason this acquisition makes so much sense is as we look to where Nvidia is going as a data center company, and more and more the workloads that our customers are wanting, they are not just single system workloads anymore,” said Boyle. “They have to be orchestrated – they are complex. And as we see a tremendous expansion into AI, one of the big gaps for our enterprise customers who do not have a strong HPC background is they need an easy way to deploy, manage and run on all of these systems.”

Just a few years ago, those AI workloads could be deployed on a single system, he said, but that is changing. “Now more and more, they are multi-system deployments. They are orchestrated. And it is what enterprises want to do, they want to deploy large language models, and they want to get into recommenders, all those great things. But they really lack the knowledge of how to run in an HPC cluster. And that's exactly where the great work that Bright's done over the years, building that software, building those management interfaces.”

The number one customer request from customers nowadays is how to make all these things work together, and that is where the Bright acquisition will help customers, said Boyle.

“It is really going to help unlock that for enterprise customers, the ability to deploy these complex workloads in a way that just makes sense for them,” he said.

Nvidia will continue to support and invest in the pure x86 capabilities of Bright software after the acquisition, said Boyle. “We see [x86 cluster management] as super important to the future because any accelerated computing workload that we have obviously has to run on a CPU as well,” he said. “But more and more of those total workflows need preprocessing of the various things in there. Some of those workloads just run on x86 and have no accelerators but are crucial to getting your data ready and everything to run your AI work.”

Bill Wagner, CEO of Bright Computing, said in a statement that it will be business as usual for Bright after the acquisition. The Amsterdam headquarters will be maintained mostly as a Bright software development office with the Bright development team becoming part of the larger Nvidia software engineering group. Wagner joins Nvidia as an executive in the company’s worldwide field operations organization, and Bright CTO Martijn de Vries will continue to lead Bright Cluster Manager development.

The current software release, v9.1 of Bright Cluster Manager, debuted in November 2020. The 9.0 release came out one year earlier.

Bright Computing launched in 2009, a spinoff of Netherlands-based ClusterVision, under the direction of Matthijs van Leeuwen, who had co-founded ClusterVision with Alex Ninaber and Arijan Sauer in 2002. ClusterVisionOS was renamed Bright Cluster Manager. ClusterVision was acquired by the Taurus Group in 2019 after being declared bankrupt.

Bright Computing’s software is widely used across industry, government and academic HPC sites, including at Microsoft, NASA, Siemens, Samsung, Johns Hopkins University and Volvo. Bright Cluster Manager for HPC received the 2015 and 2016 HPCwire Editors’ Choice Award for Best HPC Cluster Solution or Technology.

This article first appeared on sister website HPCwire.

About the author: Tiffany Trader

With over a decade’s experience covering the HPC space, Tiffany Trader is one of the preeminent voices reporting on advanced scale computing today.