Proposal Guidelines and Requirements for Allocations Larger than 10K SUs

The University has recently made a commitment to provide a major investment in research computing. As part of this investment, the University made some funds available for the fiscal year ending on June 2016. These funds have been used to:

  1. Expand aggregate storage capacity by an additional 1 petabyte. 50% of this will be dedicated to Health Sciences researchers who have pressing needs for storage and backup.
  2. Increase the number of nodes on the high throughput computing (HTC) cluster (dedicated to Health Sciences research).
  3. Expand GPU capabilities by adding 24 Titan X and 4 Tesla K40 GPUs.
  4. Add a SMP cluster for single node calculations (29 nodes, 12 cores per node). This includes nodes with large memory (~512 GB) and striped SSDs for heavy I/O calculations. This cluster will provide expanded capability and will replace old nodes on the Frank cluster. The new nodes will run applications 1.5 – 3x faster than the nodes to be retired.
  5. Add a 2688 core (96 nodes, 28 cores per node) Intel® Omni-Path interconnect MPI cluster for highly parallel applications.

This new hardware should be operational by the end of September. 12-month allocations on the SMP, MPI, and GPU clusters will be made based on proposal requests. Allocations on the HTC cluster will continue to be funded based on justification for Health Sciences research and do not require a proposal. We note also that in addition to the proposals, described below, the Center for Simulation and Modeling will continue the policy of making available 10,000 service units (roughly 10,000 CPU hours) annually to research groups that do not have major computational needs. Proposals can be submitted by a faculty member with a maximum of one proposal per group.

Guidelines for the proposal are listed below.

Proposals should be no more than 2 pages in length (Times New Roman, 12 point, 1 inch margins), describe the proposed research and justify the requested resources. E.g., if requesting time on the MPI Omni-Path cluster, the proposal should provide evidence that the software scales well over a large number of cores (here, large means > 140).

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. Proposals from existing users of the Frank cluster should include a list of all publications over the past two years that acknowledge support of these resources.

Proposals from researchers who have hitherto not been users of advanced computing resources are welcome. The consultants in the University's Center for Simulation and Modeling can assist researchers in getting started in using the Center’s computational resources.

Proposals will be accepted every four months with October 1, January 1, April 1, and July 1 deadlines for requests. Proposals should be submitted to wgj1@pitt.edu. Please use “Allocation Proposal” in the subject line. Decisions will be made within a month of those deadlines.

Groups receiving an award are ineligible to submit a new request until 12 months has passed.

GPU resources: proposals can request up to 40,000 hours/year

MPI cluster: proposals can request up to 2 million CPU (core) hours/year

SMP cluster standard nodes (256 GB ram, 256 GB SSD): proposals can request up to 400,000 CPU (core) hours/year.

SMP cluster special nodes. The cluster also has the following special nodes:

  • (a) 2 nodes (12 cores each) with 256 GB ram and 3.6 TB SSD space
  • (b) 2 nodes (12 cores each) with 512 GB ram and 3.6 TB SSD space (these will also be accessible to genomics researchers on the HTC cluster who have large memory needs)
  • (c) 1 node (12 cores) with 256 GB ram and 6 TB NVME SSD space

Users can request up to 25000 CPU (core) hours/year on (a) and/or (b). Users requesting time on these nodes should demonstrate that they have applications that do heavy I/O, or in the case of (b) demonstrate the need for more than 256 GB ram. Users can also request up to 15,000 CPU (core) hours/year on (c). In this case, the request should demonstrate that they have applications that are I/O intensive and that can greatly benefit for the use of faster NVME drives.

Other considerations: If awardees exhaust their allocation 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. Allocations will be for a period of 12 months. Unused CPU time will not be carried over to the next allocation period. Researchers who do not use their full allocation will need to provide an explanation of why the resources were not used when submitting a new proposal.