The HPC cluster management specialist fills a niche in Nvidia's software stack.
As hyper-scalers gang supercomputing clusters to support AI and other high-end automation workloads, infrastructure providers are looking for better ways to manage HPC clusters, collections of servers linked via high-speed networks.
While its acquisition of chip IP vendor Arm remains stalled in regulatory limbo, GPU leader Nvidia did manage to pull off a separate if less flashy acquisition this week, announcing a deal for HPC software specialist Bright Computing.
When asked, Nvidia said details of the transaction would not be disclosed.
Based in Amsterdam, privately-held Bright Computing was spun out from Linux integrator ClusterVision in 2009. Its software is used to provision and manage HPC and Kubernetes container clusters along with private clouds, including those running in data centers on the OpenStack cloud computing platform.
Nvidia CEO Jensen Huang promoted its acquisition of Bright Computing as a means of enabling what Huang calls “an industrial HPC era” in which HPC clusters scale to support AI workloads that dominate enterprise data centers.
The acquisition also illustrates how high-performance computing is transitioning from research laboratories to enterprise data centers and applications like factory automation.
To that end, Bright Computing has integrated its cluster management software with Nvidia’s graphics chips and GPU interconnects along with its CUDA API and, most recently, Nvidia’s DGX platform. DGX is aimed at AI training, inference and data analytics along with infrastructure support.
For Nvidia, the acquisition of Bright Computing “fills an important niche: cluster management,” said Karl Freund, an industry analyst with Cambrian-AI.
Bright Computing promotes its software as automating the management of HPC clusters consisting of up to thousands of bare-metal servers running in data centers, public and hybrid clouds and, increasingly, at the network edge. Along with Nvidia GPUs, the cluster management software supports x86 CPUs along with the Kubernetes container orchestrator.
The Linux cluster manager also supports Arm processors, complementing Nvidia’s bid for the IP vendor whose architecture is gaining traction in data center server designs, industry trackers report.
Bright Computing’s roster of about 700 customers includes Boeing, Johns Hopkins University and Siemens.
Another customer, NASA’s Jet Propulsion Laboratory, used the cluster manager to calculate trajectory corrections and landing coordinates for the Mars rover Perseverance.
This article was originally published on EE Times.
George Leopold has written about science and technology from Washington, D.C., since 1986. Besides EE Times, Leopold’s work has appeared in The New York Times, New Scientist, and other publications. He resides in Reston, Va.