Los Alamos has migrated most of its production computing to OpenCHAMIThe software supports both HPE and Dell hardware ecosystemsOpenCHAMI includes self-discovery, healing and heterogeneous orchestration capabilities
Los Alamos National Laboratory has transferred seven production clusters, totaling more than 11,000 compute nodes, to Open, Composable, Heterogeneous, Adaptable, Management Infrastructure, or OpenCHAMI, an open-source systems management stack built with other research centers and vendors.
The lab said Thursday the switch covers most of its production computing and has already reduced downtime for system upgrades.
What Is OpenCHAMI?
OpenCHAMI is cloud-native software that lets institutions use build-and-configure tools and actively manage operations and change. Its open-source code means it can be deployed with both the HPE and Dell ecosystems.
Los Alamos, HPE, the University of Bristol, the Swiss National Supercomputing Center and Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center, or NERSC, founded the community effort in 2023. Dell became a member the following year.
The founding partners set up the project under an open governance model. Their goal was to develop software that is faster, simpler and more flexible than traditional systems, which participating sites can control at scales that suit their requirements. Its capabilities include self-discovery and healing, heterogeneous orchestration and site-specific microservices.
Why Did Los Alamos Adopt OpenCHAMI?
HPC teams at Los Alamos and its partner institutions saw that demanding workloads, from data analytics and complex simulations to artificial intelligence, are difficult to run at scale in hybrid environments, according to the lab. The lab added that computing centers have historically set up their systems once and avoided changing them, with legacy tools making later upgrades harder.
OpenCHAMI gives the lab room to adapt the systems behind its mission work in modeling, simulation and AI. One capability the lab highlighted for AI work is swapping the operating system on part of a cluster’s nodes quickly.
Markus Berndt, leader of the lab’s High Performance Computing division, said the move demonstrates how well OpenCHAMI works at large sites that mix vendors’ hardware and software.
Shawn Brown, senior director of engineering for HPC and AI infrastructure at HPE, said OpenCHAMI gives customers such as Los Alamos “the flexibility to use tools that best support their mission and systems.”
In July, the National Nuclear Security Administration, HPE and NVIDIA unveiled plans to build two supercomputers, Mission and Vision, at Los Alamos. The systems will handle traditional high-performance computing and agentic AI workloads on a single infrastructure. Mission is expected to be fully deployed in 2027 and Vision in 2028.
How Is NERSC Using OpenCHAMI?
NERSC’s next-generation Doudna system will also run OpenCHAMI, with Dell supporting the deployment. Eric Roman, the center’s computational systems group lead, said the software is helping NERSC pursue the goals of its 10-year strategic plan for a computing ecosystem that is resilient, composable and agile.
Roman said that because the management software is community-driven, the center can adopt new hardware “without being locked into proprietary toolsets.” He added that NERSC can quickly repurpose compute nodes for workloads ranging from experimental data processing and large simulations to AI. That flexibility keeps maintenance outages short for the thousands of researchers at the Department of Energy’s Office of Science who use the center.
DOE selected Dell in May 2025 to build Doudna using NVIDIA‘s Vera Rubin technology. The system, named after Nobel Prize-winning biochemist Jennifer Doudna, succeeds Perlmutter as NERSC’s flagship supercomputer. It will support AI training and inference along with molecular dynamics, high-energy physics and quantum algorithm development.