Quick Start Guide: Getting an NCSA Allocation
Note: If you are new to NCSA, you might want to take a look at NCSA's New User Information page to get a quick overview of things you should know.
The following information is intended to give you a quick start to getting an allocation. All requests for allocations are submitted through the Partnerships Online Proposal System (POPS). During the submission process, you will be asked to choose a resource; therefore, it is a good idea to review the available resources beforehand.
1. Choosing a Resource
NCSA has more than 122 teraflops of computing power available to researchers. Our systems include a Dell blade system that has more than 89 teraflops of computing power, 1,200 PowerEdge 1955 compute blades, an InfiniBand interconnect and 100 terabytes of storage in a Lustre filesystem; two clusters based on Intel IA-32 and IA-64 processors that provide more than 26 teraflops of computing power; and a six-teraflop SGI Altix system. The SGI system provides a large shared-memory machine that has three terabytes of memory and 370 terabytes of online storage. The hardware infrastructure is complemented by an extensive infrastructure of open-source and commercial software. When applying for time on NCSA machines, give careful thought to the appropriateness of the requested resource(s).
Allocations are awarded based on the number of service units (SUs) requested. The number of SUs that a computer job is charged usually represents the number of CPU hours accumulated over all the processors that the job uses. SUs vary per resource. Helpful information will be found within the Technical Summary page for each compute resource. Start at the NCSA Compute Resource page (linked below) to learn more about each machine and the charging on the available systems. Information on SUs and NCSA's charging algorithm is also available.
For additional information, contact Consulting.
2. Writing a Proposal
For a Development Allocation, you are required to submit a 250-word abstract and attach a resume or CV. No full proposal is necessary. It may take up to four weeks to process this request. The PI will receive an e-mail confirmation regarding the award decision.
For MRAC and LRAC allocations, you can find details for writing a proposal to use any NCSA resources in the TeraGrid: NSF Resource Allocation Policies. See various samples of successful, well-written MRAC (NCSA, TeraGrid) and LRAC ( NCSA, TeraGrid) proposals. Reviewers of successful proposals particularly noted that:
- The proposals summarized the research in the context of the current state-of-the-art, outlined the computational algorithms to be used, and related these algorithms to the research subsections.
- The reviewers were provided sufficient information but not overwhelmed by details.
- Justification for the service unit (SU) request was clear and closely coupled to the computational experiments. If the committee will need to reduce the original request, it may be done rationally with minimum disruption to the investigator.
- Previous allocation awards were summarized and related to the request in the current proposal.
3. Submitting a Proposal
Proposals for computing allocations must be submitted electronically via POPS, including allocation requests for the TeraGrid systems. If you submit multiple proposals, you should only do so if the projects are in different fields of science. If submitting more than one proposal, you must attach a letter to each one stating that they are in different fields of science and explain in detail why they can not be combined. Without sufficient explanation, the review board may consolidate the proposals and issue a single award.
High-Performance Computing (HPC) allocations are awarded based on the number of Service Units (SUs) requested. The number of SUs that a computer job is charged usually represents the number of CPU hours accumulated over all the processors that the job uses.
