Frequently Asked Questions (FAQs) for NSF 14-536, High Performance Computing System Acquisition: Building a More Inclusive Computing Environment for Science and Engineering

Yes.

The project description should make a compelling case that if only the new computational resource/capability or the upgrade of an existing system to encompass new capabilities were to be funded, that the impact of the separation would be minimal. The budget pages should clearly separate the two resources so that it is easy to see how to proceed.

Attention is drawn to https://www.nsf.gov/pubs/policydocs/pappguide/nsf14001/gpg_sigchanges.jsp. The recent changes to the NSF GPG were fairly substantial. You are strongly encouraged to closely consider this updated guidance in every detail before submitting a proposal.

Prospective PIs should discuss this issue with their sponsored project office, which should be
able to provide the necessary guidelines to properly frame the proposal. In addition, PIs may want
to reference NSF’s guidance on preparing the Facilities, Equipment and Other Resources section
of the proposal. That guidance is found in Chapter II.C.2.i of the GPG
(https://www.nsf.gov/pubs/policydocs/pappguide/nsf14001/gpg_2.jsp#IIC2i).

Yes. NSF would be prepared to fund one part of a combined proposal if the recommendations from
the panel were sufficiently compelling. It is important for PIs to consider how their proposals are
constructed to allow the panel to make those recommendations in a clear and compelling fashion. In
addition, the budgets must be constructed to make the separation straightforward.

Notably, while in past solicitations benchmarks have played an important role, for this solicitation, given the emphasis on nontraditional HPC as well as possible “cloud-like” computation solutions, and the potential breadth in targeted scientific impacts, NSF has opted not to require a set of NSF-provided benchmarks. Nonetheless, NSF will still require that each proposer provide a convincing demonstration, with hard data, that their system will perform as described in their proposal.

Attention is drawn to the requirement that the proposer present the acquisition implementation, project management, and mitigation of acquisition risks in the context of the anticipated cooperative-agreement award type. Potentially, this presentation could include a proposed project manager who would provide expertise to the submitting organization in the management of awards of this scale.

Attention is drawn to the solicitation’s important requirement that the hardware acquisition must provide a high degree of stability and usability by no later than January, 2016. Acquisition and deployment of the full system should be finished before the end of FY16, i.e., September 2016. The end of FY16 should be interpreted as a not-to-exceed date. It is perfectly acceptable to begin full deployment earlier, or to begin partial deployment earlier and ramp up. Again, this will be negotiated in the cooperative agreement, as it could be dependent on the nature of the resource proposed and funded. The number of years that the proposed system will be deployed can vary with the nature of the proposed resource, but in most cases it is anticipated to be part of XD for at least 4 years.

Attention is drawn to the requirement that user support and operating costs per annum will be no larger than 20% of the acquisition cost for each deployed HPC system. These costs will be covered in a separate funding action and the size will depend on the nature of the resource deployed.

Awards anticipate that 90% of the capacity/capability of the delivered systems will be available to the open science community through the national allocation process, XRAC.

This will involve a subset of the seven bullets listed in Section II, Program Description, of the solicitation. A competitive High Performance Computing and Data (HPC&D) proposal must address one or more of those seven items. As stated thereafter in the solicitation, in FY 2014 NSF is interested in receiving innovative proposals for production XD computational and data resources capable of complementing recent NSF HPC investments.