Leadership-Class Computing Facility

Discovery at the frontiers of science and engineering requires computing power at a massive scale, sustained over a long period of time. To reach this goal, the National Science Foundation (NSF) is funding the design of a Leadership-Class Computing Facility through its Major Research Equipment and Facilities Construction (MREFC) process, with an operations model on a multi-decade scale.

NSF is completing the planning process for a Leadership-Class Computing Facility (LCCF), slated to become part of NSF’s Major Research Equipment and Facilities Construction (MREFC) program. This facility will support a ten-fold or more time-to-solution performance improvement over Frontera, the current NSF leadership system. If ultimately awarded for construction, the LCCF will represent an enormous step forward in both the scale and the long-term sustainability of advanced computing resources available to the open science community.

The MREFC process has three design review gates – conceptual, preliminary, and final – that must be passed with increasing levels of scrutiny on design and cost. Construction is planned to begin in March 2024.

The first milestone, in the summer of 2020, led to a conceptual design approved by NSF’s Large Facilities Office and evaluated as part of the MREFC process. In early 2022, TACC and NSF presented the preliminary design to a peer review panel, who strongly recommended that the project move to the Final Design stage. The final design package completed peer review in early 2023, with a strong recommendation to proceed to construction. The NSF is conducting its administrative review process on the design package now.

The facility is planned for in late 2026 to ensure the continuous availability of a large-scale computing resource for the NSF research community. A unique feature of MREFC projects is that the facility is guaranteed a 10-year initial operating period that may be renewed indefinitely. This feature is especially valuable for other NSF large facilities and long-term science projects for which the prospect of changing supercomputing centers every five years as typical cyberinfrastructure grants expire is a significant hurdle.

The conceptual and preliminary designs included the items below, along with an initial 10-year operating plan: