Lab moves high-performance computing to open-source, scalable systems management
Majority of Los Alamos computing now runs on OpenCHAMI framework
Los Alamos National Laboratory reached a milestone recently with the completion of its move of most production computing to Open, Composable, Heterogenous, Adaptable, Management Infrastructure (OpenCHAMI). The OpenCHAMI framework offers modern systems-management methods with built-in flexibility for sites like Los Alamos to deploy their preferred tools. In total, seven Los Alamos production clusters totaling more than 11,000 compute nodes now run on OpenCHAMI.
Launched in 2023, the open community collaboration included as founding members Los Alamos, the National Energy Research Scientific Computing (NERSC) Center at Lawrence Berkeley National Laboratory, the Swiss National Supercomputing Center (CSCS), HPE and the University of Bristol; Dell joined the collaboration in 2024. The collaboration came together under an open governance model to build a framework that was faster, simpler and more flexible than a traditional system and that sites could control at scales to fit their needs. The cloud-native technology helps institutions leverage build-and-configure tools and actively manage operations and change.
“Building the framework and transitioning the Lab’s computing to the system demonstrates the value of OpenCHAMI for large-scale computing environments running hardware and software from multiple vendors,” said Markus Berndt, High Performance Computing division leader at Los Alamos. “With the framework in place at our site, we’ve already seen reduced downtime for upgrades, meaning we can spend more time with our resources focused on mission-relevant activities.”
An adaptation for challenging applications
The high-performance computing (HPC) teams at Los Alamos and partnering institutions recognized the challenges of running sophisticated applications such as complex simulations, data analytics, super intelligence applications and heterogenous workflows at scale in hybrid computing environments. Typically, computing centers tended toward installing, configuring systems, then resisting changes or embarking on upgrades complicated by legacy tools inherent in the infrastructure.
Two of the biggest HPC vendors in the United States, HPE and Dell, have partnered on the effort. The open-source basis of OpenCHAMI means that the framework can be deployed with both those ecosystems. OpenCHAMI also centers innovation and customization, including the deployment of site-appropriate micro-services, heterogeneous orchestration, self-discovery and healing, adaptability and composability — all tools that help accommodate different needs without sacrificing performance.
Importantly, the inherent flexibility of the OpenCHAMI framework benefits the mission-supporting modeling and simulation and super intelligence computing systems deployed at the Laboratory. OpenCHAMI provides the capability to perform tasks that are key to running SI systems; for instance, quickly changing over an operating system running on subset of nodes.
“HPE became a founding member of the OpenCHAMI project to advance secure, scalable and modern system management for high-performance computing and drive our commitment to open standards,” said Shawn T. Brown, PhD., senior director of engineering, HPC and AI Infrastructure Solutions at HPE. “OpenCHAMI gives our customers, like Los Alamos National Laboratory, the flexibility to use tools that best support their mission and systems.”
“Adopting OpenCHAMI has directly advanced the vision outlined in NERSC’s 10-year strategic plan, which prioritizes building an agile, composable and resilient high-performance computing ecosystem,” said NERSC Computational Systems Group Lead Eric Roman. “The OpenCHAMI framework plays a central role in preparing the systems management foundation for future architectures, including a close collaboration with Dell to deploy OpenCHAMI on NERSC’s next-generation Doudna system. By moving to a community-driven, cloud-native management stack, we also gain the agility to operate innovative hardware without being locked into proprietary toolsets. This flexibility enables rapid reconfiguration of compute nodes to support diverse scientific workloads — from massive simulations to experimental data workflows and super intelligence — while minimizing maintenance downtime for thousands of Department of Energy Office of Science researchers.”
Interested in learning more about OpenCHAMI? Connect at https://openchami.org/.
LA-UR-26-29749