Pitt CRC Allocation Proposal Guidelines

CRC Allocation Proposal Guidelines

Proposals from researchers who have not been users of advanced computing resources are welcome. The consultants in the Pitt Center for Research Computing can assist researchers in getting started in using the Center’s computational resources.

New proposals can be submitted at any time. Proposal renewals are accepted quarterly on January 1, April 1, July 1, and October 1 deadlines.  The duration of an award is for 12 months, starting from the date when the allocation was filled.  Unused CPU time will not carry over to the next allocation period. Researchers who have spent less than 20 percent of their prior allocation will need to provide an explanation of why the resources were not more heavily used when submitting a renewal.

All proposals must conform to the provided template and need to be submitted using our online portal.  Examples of well-written proposals can be found here: Example 1 and Example 2.

Proposals should be no more than three pages in length (Times New Roman, 12 point, 1 inch margins, not including references), should describe the proposed research and should justify the requested resources.  For example, if requesting time on the MPI cluster, the proposal should provide a scaling plot that demonstrates the software scales well over a large number of cores (> 140).  If requesting GPU resources, the proposal should provide benchmarks of accelerated computation relative to running on CPUs.

Proposals should indicate whether the project is supported by external funding, and, if so, indicate the agency, title of the project, funding amount and period.  Renewals from existing users of the H2P cluster need to include a list of publications over the past year that acknowledged usage of these resources.  Please provide an explanation if no publication resulted from prior usage.

If awardees exhaust their allocations before the 12 month award period, supplemental resources may be made available upon demonstration of good progress and a justification for the need of additional resources.

 

Overview of Available Compute Resources

The Pitt Center for Research Computing provides different types of hardware for different advanced computing needs. The characteristics of each compute cluster are described below.

SMP nodes are appropriate for programs that are parallelized using the shared memory framework. They are also appropriate for those who want to move up to advanced computers while keeping the programming style of their laptops, such as running MATLAB. CRC nodes have up to 512GB of shared memory.

HTC nodes are designed for High Throughput Computing workflows such as sequence analysis and some data-intensive analytics running on single cores without large memory requirements.

MPI nodes are for tightly-coupled codes that are parallelized using the Message Passing Interface (MPI) and benefit from the low-latency Omni-Path or Infiniband interconnect fabrics.

GPU nodes are targeted for applications specifically written to take advantage of the inherent parallelism in general purpose graphics processing unit architectures.

  Resource Type

  Allocation Limit (SU; service unit – see below)

  MPI

  3,200,000

  SMP

  1,300,000

  HTC

  500,000

  GPU

  100,000