Guidelines for Preparing a Proposal for the Seawulf Cluster
It is important that your proposal contains information that lets the reviewers determine whether your plans are well suited to Seawulf, and the proper size of the allocation that is needed to perform your proposed research. The proposal will be reviewed to ensure that you have properly estimated the resources that your project will require, and determine if additional support of software are needed. For graduate student applicant, the faculty advisor should prepare the proposal.
- Briefly explain the background of the project in approximately one page of the proposal. Please describe the project in terms that can be understood by a broader scientific audience and not just an expert in your field.
- What are the overall goals? Provide sufficient information but do not overwhelm reviewers with details.
- How does the computational aspect fit into the broader research plan?
- Is the project funded (NSF, NIH, etc), or is a proposal planned?
- Explain the computational plan in sufficient detail so that the reviewers can determine if the plan is solid and the resource request is properly calculated and justified. Provide about one page of detail for this section.
- What is your experience using a computer like Seawulf? Will you need help?
- Which software will be needed? Is it freely available?
- What is your experience running this software? Will you need help?
- How many runs are needed and why? For example, if 20 runs are requested, why is 10 not sufficient?
- Why is this length of run needed? For example, why is 20ns of molecular dynamics needed instead of 10ns? The length should be justified by existing data, or you should explain your rationale in some detail.
- How many processors will be used for each run? What is the scaling efficiency? Is this the right number to be used? (If this data is not available, please request a smaller allocation to obtain the data, or work with the Seawulf staff to get benchmark data).
- What formula was used to obtain the total CPU hours requested? This should include number of runs, length of each run, CPU hours required for each, and a clear calculation that shows how the total amount requested was obtained using these values.
- In the absence of clear data justifying the size of the request, a small request may be approved for projects of scientific value. This small request can be used to gather the data neede to assess a larger request.
- If you envision your computational needs growing in the future, what are your plans to obtain resources at other locations such as the large national supercomputer centers? The Seawulf staff may be able to help you identify appropriate centers. For example, the NSF-funded centers have a web page explaining the proposal process at http://www.ci-partnership.org/Allocations/.
- Make sure that you fill out all of the sections on the allocation form- do not leave any blank.
- Investigators should attempt to submit a minimum number of proposals. Investigators should only submit multiple proposals if the projects are in different fields of science or if a new project is undertaken after an allocation has been made for a previous project.
- If you have previously had an allocation on Seawulf, you should summarize the results that were obtained in your new proposal. A final report as described on the Seawulf web site is required for all completed proposals. Completion of the report will be a condition for consideration of any requested renewal..