Research Computing Consulting Policies
Expert HPC Consulting...
Research support services are provided by research computing consultants working closely and collaboratively with university researchers. Two types of consulting services "open," or short-term, and "extended," or long-term, consulting are offered and defined below. If a PI indicates that consulting is needed, the consulting should take place before CPU time is allocated. Consequently, consulting is an (optional) first step in a new project being approved for the HPC environment. The distinction between open and extended consulting is only one of time (see below for details). Often either the consultant or the HPC user will be able to determine in advance which kind of consultation is required for a given problem. Consequently, the researcher and consultant should meet early in the consulting project planning process to mutually agree on the parameters of a project.
“Open consulting” is a short-term consulting service (8 hours or less) available to all university researchers. Open consulting projects are offered to help develop or refine an existing application for extended consulting services (see below) or to provide general short-term consulting services. Examples of open consulting services include assistance or advice on building codes, providing letters of support for grant proposals, determining whether existing software is suitable for solving a new problem, or basic advice on parallel programming techniques. In addition, open consulting is available for simple database design and the addressing of a researcher’s specific storage needs. These services all involve research consultants sharing expertise with researchers, but open consulting services may also include research consultants’ providing services to researchers such as database query optimization, code optimization, parallelization, or porting a code to the HPC environment. In addition, research consultants are always available to meet with a researcher in order to discuss pursuing new avenues of funding.
“Extended consulting” is consulting that requires significantly more time than open consulting. Because of the larger time commitment, the availability of extended consulting is dependent on the availability of a research consultant with suitable expertise in the problem. The design and/or optimization of large databases typically would be an instance of extended consulting. Further examples of extended consulting services include parallelization of large research codes, the development and/or implementation of parallel algorithms, or development of novel simulation techniques and most of the research application development services. Extended consulting of up to 160 hours is available at no charge to the researcher, pending the availability of the consultant and approval of the consulting project application. Longer term consulting is also available, but fees for the consulting time do apply. Fees may also apply for shorter projects in cases where IT must supplement its staff or use resources outside of the research-computing group in order to provide the service. Consulting projects requiring additional full or part-time (soft-money) staff will require approval by the VP for IT and will always be conditionally approved until the positions are approved and filled. The possibility of IT sharing the cost of or managing additional personnel will be considered on a case-by-case basis. In order to arrange extended consulting, whether for a fee or not, the researcher should meet with a consultant to determine the feasibility of the project and complete the online application process.
CONSULTING PROJECT POLICIES
A typical open consulting project will not involve extensive code development or the development of novel algorithms by the consultant. Projects involving such services are indeed research collaborations, and the consultant will receive due credit and co-authorship on these collaborations. Similarly, the consultants will not run applications or perform data analysis for users except for instructional purposes. A typical project will involve the consultant’s recommending specific software or methods and/or instructing the user so that they may implement the suggested approach for themselves if desired. For typical projects, an acknowledgement of the Academic Computing Division in any presentations or publications is appropriate. When a researcher enters into a consulting agreement, it is implicitly understood that the researcher will inform IT (either voluntarily or by answering surveys) of published or presented work resulting from the use of HPC resources. Likewise, after a consulting project is completed, researchers should provide a short summary of the results of the collaboration including a list of researchers involved as well as any grants, publications, online resources, and presentations that resulted from the project. This information will be used for an online database/portfolio of projects supported by IT’s research computing group. This information will also be used to determine effectiveness of particular projects and for metrics that may be used to assess requests for further funding for new HPC resources.
As a rule of thumb, open consulting is limited to a few meetings or roughly 8 hours of consulting time. Whether for a fee or not, extended consulting projects are sufficiently complicated that a resource allocation policy will be necessary to ensure that IT’s limited research-computing resources are used both in a fair manner and in one that is consistent with the strategic plan of the university – with particular attention to the university’s research goals and IT’s strategic plan for research computing. One of the primary components of the resource allocation policy will be the required application and review process for extended consulting projects. Resources will be allocated on a project basis to a single PI from the University of Louisville. Consequently, visiting researchers, graduate students, postdoctoral students and undergraduates should have a university faculty member assume responsibility for the project and submit a request on their behalf.
The review process will be similar to that in place for the university’s IRIG (Intramural Research Incentive Grants) program. The primary criteria for application approval will be the academic soundness and feasibility of the project along with the availability of a consultant with suitable expertise. The immediate timeline of the project (e.g. an imminent proposal deadline) and the consulting time requested (shorter projects will be favored) may lead to an increase in priority. Likewise, alignment with the university’s strategic research goals and the potential to increase the funding of the university and/or the academic computing division will be considered when reviewing applications and allocating resources. To manage cases where contention for resources occurs the governance committee will play an ongoing role in the review and approval process by either directly reviewing or forming a subcommittee to review requests that require existing projects to be reprioritized and/or new projects to be postponed or denied.
APPLICATION DEVELOPMENT SERVICES
For extended consulting projects, research computing consultants are available for the design and development of new research applications. We provide guidance in all aspects of research application development: hardware and software selection, project planning, programming tools, techniques, and languages. We will also assist with debugging of parallel applications and the modification of existing research codes:
- Parallel or serial optimization of an existing application including algorithm development or modification (e.g., incorporation or implementation of new or advanced numerical methods or libraries).
- Porting an existing application to the CRC
- Development of a parallel code from an existing serial code or algorithm.
- Incorporation of new database management or storage methods.
- Incorporation of new visualization methods.
Contact us at email@example.com with any questions or for additional information.