Tools and Resources
Tools and Resources
How To Apply for Computational Time
Mar 12, 2025. Guidelines
Click below to download a copy of the document.
Purpose/Description
Access to high-performance computing (HPC) resources is essential for advancing scientific research and innovation. Applications for computational time at major U.S. supercomputing centers, for supercomputers such as Perlmutter, Frontier, and those at other national facilities, follow a structured process that includes writing a proposal, meeting eligibility criteria, and adhering to specific submission guidelines.
This guideline briefly describes these supercomputing centers and details the application process, including proposal writing, submission requirements, and deadlines. It also explores different allocation call types and scaling analysis's role in demonstrating that a code is ready for HPC execution and can effectively utilize large-scale computing resources. Additionally, we highlight available grants and funding opportunities to support HPC-driven research. Finally, we present best practices for writing strong proposals and ensuring successful submissions.
Table of Contents
- Major Supercomputing Centers in the U.S
- How to Apply for Computational Time
- Grants and Funding Opportunities
- General Best Practices for Proposal Writing and Submission
- Conclusion
1. Major Supercomputing Centers in the U.S
The Department of Energy (DOE) has several supercomputing user facilities at our national laboratories, including the first exascale computer in the world. Many of these supercomputers are available to industry researchers, academia, and the public through the application process outlined in this guideline document.
DOE currently operates three exascale computers: 1) El Capitan at Lawrence Livermore National Laboratory (LLNL)*; 2) Frontier at the Oak Ridge Leadership Computing Facility (OLCF); and 3) Aurora at the Argonne Leadership Computing Facility (ALCF).
*The National Nuclear Security Administration’s (NNSA’s) El Capitan is officially the world’s most powerful supercomputer and the first exascale machine dedicated to national security.
See below for more information about these centers and recommended best practices to apply to them for computational time.
-
-
- Location: Lawrence Berkeley National Laboratory (LBNL), Berkeley, California.
- Supercomputers: Perlmutter and El Capitan
- Focus: Perlmutter is designed to support scientific research across various domains, emphasizing energy science, climate modeling, high-energy physics, and fusion research. El Capitan, an exascale supercomputer, supports unclassified projects in energy security, climate change, cancer drug discovery, and other areas of public interest.
- Usage: NERSC typically prioritizes projects that advance scientific knowledge in areas aligned with U.S. Department of Energy (DOE) missions.
-
-
-
- Location: Oak Ridge National Laboratory (ORNL), Oak Ridge, Tennessee.
- Supercomputer: Frontier
- Focus: Frontier is the world's first exascale supercomputer, supporting scientific research in materials science, biological science, fusion energy, and quantum computing.
- Usage: Frontier provides a degree of computing performance otherwise generally unavailable for large-scale simulations and data analytics, offering access to state-of-the-art computing power for projects requiring massive parallelization.
-
-
- Other Major Supercomputing Centers
-
-
- Argonne Leadership Computing Facility (ALCF) at Argonne National Laboratory, Argonne, Illinois: Includes Aurora, one of the fastest supercomputers, designed for data-intensive simulations.
- Los Alamos National Laboratory (LANL), Los Alamos, New Mexico: Strong nuclear and materials science concentration. This includes the newly inaugurated Venado supercomputer, which specializes in AI-intensive applications.
- Texas Advanced Computing Center (TACC), Austin, Texas: Frontera and Stampede3 for general scientific simulation.
-
2. How to Apply for Computational Time
To request computational time, you must submit a scientific proposal containing scientific needs and probable outcomes to the computational facility (i.e., supercomputer center). Many supercomputer centers, such as NERSC and OLCF, serve only U.S. institutions. Note that sometimes international collaborators are accepted. In most cases, following a given template and adhering to a maximum number of pages is mandatory.
a. Writing and Submitting Proposals
General Types of Proposals:
-
-
- Regular Proposals: This is the typical application for computational time, often reviewed annually or bi-annually.
- Director's Discretionary Time Proposals: Use this application type to request that the center directors allocate time to specific projects.
Note: Specific proposal templates and submission formats may vary by supercomputer center. Always consult the center's guidelines before submitting proposals to ensure compliance.
-
b. General Proposal Guidelines
-
-
- Executive Summary: Clearly define the scientific question, methodology, and anticipated impact. Explain why the requested computational resources are needed and how they will enable the project.
- Technical Approach: Describe the computational methodology, including but not limited to issues of parallelism, scaling, and input-output data handling. Explain how the proposed work leverages the facility capability, including, for example, GPU acceleration or exascale processing.
- Scientific and Broader Impact: State the anticipated outcomes and impacts of your research in general, scientific, and societal terms.
- Computational Resources Required: Estimate the core hours or node hours required. Additionally, outline other resource requirements such as memory, storage, and GPUs. Specify any dependency on special hardware, such as high-memory nodes.
- Prior Experience: Discuss previous experience with similar computational models or HPC. If applicable, indicate past successes in winning computational time or managing large-scale simulations. Note that the principal investigator (PI) usually has previous experience with HPC resources, while other team members, often in earlier career stages, are designated as co-PIs or computational PIs and are the primary users of the computational resources.
- Team and Collaboration: Provide the names of team members and collaborators and their roles. Mention cross-institutional or interdisciplinary work.
- Timeline and Milestones: Divide the project into phases. For each phase, specify the expected milestones, deliverables, and timelines. Justify the time requested in each phase. Provide a Gantt-type diagram when/if possible.
-
c. Submission Deadlines
-
-
- Annual or Semi-Annual Deadlines: Most facilities have set deadlines, usually in Spring and Fall. For example, NERSC and OLCF have specific deadlines for proposal submission, often in October and June, respectively.
- Director's Discretionary Time Deadline: Such allocation can sometimes be proposed independently of the regular deadlines, generally takes less time, and requires exceptional scientific merit.
-
Check the call for proposals, submission guidelines, and the exact deadlines directly at the website of the respective supercomputer center:
-
-
- NERSC: https://www.nersc.gov
- OLCF: https://www.olcf.ornl.gov
- ALCF: https://www.alcf.anl.gov
- LANL: https://www.lanl.gov
- TACC: https://tacc.utexas.edu
-
d. Types of Calls
Before applying for supercomputer computational time, you must consider the different call types, each suited for different research needs. The most common call types are listed below:
-
-
- Regular calls, which apply to regular projects needing only moderate resources.
- Exascale calls, which deal with large-scale simulations requiring the latest in high-performance computing resources.
- Benchmarking calls, which test and optimize code performance on supercomputing architectures in preparation for high efficiency at full-scale runs.
- AI-intensive calls, which offer dedicated resources for data-intensive and machine learning workloads for projects related to artificial intelligence.
-
Each call type has its unique requirements and evaluation criteria. Hence, researchers should carefully align their proposals with the appropriate category to maximize their chances of success.
e. Scaling Analysis
As part of their computational resources requirement, applicants are typically asked to provide strong and weak code scalings that determine, for example, the optimal number of nodes or CPU cores they will need. This analysis evaluates how efficiently a parallel application utilizes increasing computational resources to assess performance improvements and identify bottlenecks.
Refer to this Princeton Research Computing webpage and remember that you must understand your code to get the most out of it. Benchmarking calls are separate calls for proposals to allocate resources for benchmarking rather than for fundamental science projects. For benchmarking calls, any computer resources provided are intended explicitly for benchmarking calculations. This includes strong and weak scaling tests of your code under the specific conditions defined by your computational and physical problem.
3. Grants and Funding Opportunities
Many supercomputing facilities require projects that are associated with funded research, including but not limited to the following:
-
- DOE Grants: For Frontier or NERSC, alignment with DOE-funded research is often required.
- National Science Foundation (NSF) Grants: Research proposals that receive NSF funding.
- Independent Grants: A few supercomputing centers, such as TACC, offer independent grants for beginning researchers in a career or focused research topics like fusion research.
Please confirm eligibility and funding recommendations for each facility and, when necessary, submit requests for funding to these sources.
4. General Best Practices for Proposal Writing and Submission
-
- Be clear and concise: Avoid technical terms that overwhelm reviewers outside your specialty.
- Computational resource justification: Clearly explain why the requested resources are necessary and provide evidence that the project can effectively deploy the allocated computational time.
- Scientific outcomes: Reviewers expect to see that the project will have quantifiable scientific outcomes.
- Engage early with facility staff: Many centers offer support, consultation, and guidance for proposals at the time of submission.
Conclusion
Proposals for computational time at major U.S. supercomputing centers require thorough preparation and proposal writing, a clear understanding of resource allocation procedures, and careful attention to deadlines. To be competitive, a proposal must align with the center’s mission, demonstrate the project's scientific impact, and provide a compelling justification for the requested HPC resources. A well-crafted proposal can grant access to world-class computing power and enable groundbreaking scientific discoveries. Best of luck!