Executive Summary

In 2022, Duke conducted a comprehensive study of Research IT Needs, across all its non-clinical scholarly domains. Following the release of the December 2022 summary report, Duke’s Vice President for Information Technology, Vice President for Research and Innovation and Vice Provost for Library Affairs teamed up with others to sponsor Phase 2 of the effort—developing service proposals to meet the Phase 1 needs expressed by faculty, which encompass much more than IT.

From January-May 2023, six cross-functional staff teams—each with faculty representation—drew up 39 proposed services to address Phase 1 expressed needs. After further service refinement and faculty and sponsor feedback, twelve services are recommended for implementation and they comprise three overarching service clusters, illustrated below, with the relative service priority of each enumerated:

Descriptions of the 12 service proposals in a table

Together these three overarching service clusters and their twelve services reflect a coordinated and integrated research support program across the Office for Research and Innovation (ORI), Office of Information Technology (OIT), Duke University Libraries (DUL) and others, in conjunction with Schools. Five of the twelve proposed service are already being actively pursued by service partners.

These services advance to Phase 3 which will focus on the funding approaches to implement the services (July-September 2023). Phase 3 will be led by financial experts and is likely to involve a multimillion-dollar funding increment on top of the indirect cost recoveries already in place today that support Duke’s research endeavor (in excess of $300M). The funding approaches identified in the Phase 3 process are expected to vary for different services, from university (allocation) funded, to direct-to-grant funded, to overhead (indirect cost recovery) funded. Philanthropy may also be relevant and in some cases a service might be funded in part or in full by internal budget reallocation at the service provider level (but this is assumed to be the exception rather than the norm).

Beyond these twelve services, other proposed services were compelling and may represent targets for future funding or local (School-specific) pilots. As one survey respondent put it, “the solutions are so refined that they ALL sound nearly equally compelling.” Of note, two pilots are already being pursued with Engineering for services that were not advanced to Phase 3 because their need was more localized.

 

Background and Context for Phase 2

From February-November 2022, Duke’s Information Technology Advisory Council (ITAC) undertook an assessment of Research IT Needs at Duke. The assessment invited participation from 37 faculty and 2 research/teaching staff1 who were identified by deans and drawn from non-clinical scholarly domains. Their input was synthesized into six major findings and ten recommendations reflecting areas of common, perceived need. It is important to note that various findings and recommendations extended beyond the IT domain to encompass research support more generally.

The result of the assessment was presented to Academic Council on December 1, and a summary report2 was simultaneously released to document the process and outcomes. Soon after, the Vice President for Information Technology, Vice President for Research and Innovation and Vice Provost for Library Affairs joined together as the three primary sponsors in establishing Phase 2 of the Research IT Needs assessment. The Phase 2 work was designed to be carried out by six cross- functional teams, whose purpose was to identify, develop and prioritize service proposals or projects that would be responsive to the recommendations identified by researchers in the Phase 1 effort.

 

Phase 2 Working Group Formation and Service Proposal Development and Consolidation

In January 2023, the primary sponsors identified charges and membership for six working groups (teams), one for each of the Phase 1 finding in areas of People, Process/Structure and Technology. The primary sponsors solicited eight other leaders (deans and other administrative executives), each of whom joined in sponsoring one of the Phase 2 working groups. The teams were identified as Groups A-F, corresponding to the finding from Phase 1 on which they were tasked to work.

Each team was largely comprised of staff drawn from ORI, DUL, OIT, and DHTS (Duke Health Technology Solutions), with others. The concentration of membership from these units was acknowledgement that those groups were likely providers of future services and solutions arising from Phase 2. Each group also included at least two faculty champions and other faculty consultees who were selected to monitor the working group’s emerging service and project proposals to ensure the process remained researcher-centric and solutions were responsive to the faculty needs. A total of 55 individuals—26 staff and 29 faculty—were invited to participate in the six Phase 2 working groups, serving in distinct roles as Leads, Members, Faculty Champions, Faculty Consultees, Consulting Experts, and Staff Facilitators. Six of the faculty invited to engage in Phase 2 were active participants from Phase 1 and their inclusion provided a feedback loop between the phases. Other Phase 2 faculty were drawn from ITAC (9) to retain linkage to the body that initiated and oversaw Phase 1, and the balance (14) were drawn from faculty at large in response to Phase 1 feedback from deans, the provost, Academic Council, and others. (See Appendix A for membership and charges.)

 

1 For simplicity, this report will refer to both the 39 faculty and 2 research/teaching staff participants in Phase 1 as faculty.

2 The full report from Phase 1 is available at https://duke.is/72sjn.

 

Working groups met with sponsors in February to receive their charge, then each group met weekly over the next ten weeks to review Phase 1 findings, evaluate existing resources and associated gaps, and identify solutions they believed would be responsive to the needs expressed in Phase 1. Meeting frequency for Phase 2 participants was based on role, with each team’s 3-4 leads called on more extensively (weekly) than members (bi-weekly) or faculty champions (3 times in the 10 weeks).

Faculty had the option to engage as extensively as they wished, some choosing to participate deeply and others electing limited participation. Consultees /experts were called upon only as needed.

By early April, the six working groups had identified 39 potential services they believed could help meet the researcher needs expressed in Phase 1. These services are described in Appendix B. On April 6 a poster session was organized for all six groups, with sponsors and participants from both phases of the project invited. More than 50 individuals attended, about half representing Phase 2 staff leads, members, consulting experts and facilitators, and the other half consisting of faculty and sponsors. The poster session stimulated conversation across the various working groups and faculty provided feedback regarding the 39 proposed services (see Appendix C). As a result, ten service proposals were consolidated into other, similar proposals, leaving 29 distinct services / proposals for further consideration. (Appendix D details the consolidation process.)

During April and early May, a readout for each working group was held with its sponsors to discuss and refine proposed services. The status of the Phase 2 effort also was reviewed with several groups at the end of the academic year: with Deans Cabinet on April 24, with the Faculty Subcommittee of EROC (Executive Research Oversight Committee) on May 2 and with ITAC on May 11.

In May, poster session feedback, sponsor readouts, and guidance from a faculty expert in survey design were used to further pare the 29 distinct services/projects down the 21 service proposals that were most responsive to the overall needs of researchers as expressed in Phase 1. (See Appendix D.)

Prioritization Process and Resulting Categories of Recommended Outcomes

Proposed services were rated by faculty and sponsors, then graphed. As the following conceptual graph illustrates, they fell into five broad categories with three associated outcomes: twelve services are recommended to advance to Phase 3 for funding strategy development; two services require further evaluation; and at least five services may represent partnering opportunities with Schools.

 

A graph of faculty compared to sponsor strategic priorities

Appendix E provides an actual graph showing how each distinct Phase 2 service aligns with the different categories (or falls outside them), based on faculty and sponsor ratings, as well as cost. Categories and related services are described next, with estimated costs noted in shorthand ($-$$$).

 

Faculty Top Quartile Priorities: Faculty from Phases 1&2 (n=58) were surveyed regarding proposed services. Response rate was an impressive 67% overall and based on mean faculty ratings3, five top quartile services emerged as highest priority (each with an average score >2.2 score on 3-point scale). The faculty survey responses and write-in comments appear in Appendix F.

  1. Add 15-20 FTEs spanning Libraries, ORI, OIT and Schools to enable and improve new categories of research support and provide more consistent offerings to units. ($$$)
  2. Devise tools to manage data over its life cycle, understand storage cost, and clarify where data reside. Provide storage capacity to meet 80% of active research project need. ($$)
  3. Enhance VM provisions (processing / memory) in the Duke Compute Cluster (DCC)

that are provided to researchers; extend access to graduate (PhD) students and postdocs. ($)

  1. Provide storage flexibility to meet differing research needs (secure + public access) that are

compliant w/ regulations for storage retention. ($$$)

  1. Develop training programs for faculty and students (grad and undergrad) and ensure IT personnel are well trained on research support services. ($$)

Sponsor Additional Priorities: Sponsors next rated strategic impact of the proposed services4 and four more service priorities resulted. Each garnering an average sponsor rating >4.5 and was also highly rated by faculty (scoring above the median).

  1. Better support AI/ML and other research through GPU capacity in the DCC, similar to the

DCC’s on-demand CPUs access (shared and scavenged). ($)

  1. Use a risk-based approach to establish security and compliance expectations at a project level, based on regulations, risk, and data classification; include guidance for how exceptions can be requested. ($$)
  2. Build cross-department virtual teams, to better support personnel across Schools and in ORI, OIT, and Libraries, using 1-3 FTEs to manage, develop and support the personnel. ($$)
  3. Institute protected enclaves to encapsulate individual project data with requisite security protections; enable authorized access/data movement based on the project circumstances. ($$)

High Impact/Low-Cost Priorities: Finally, very rough cost estimate ranges were developed by working group staff, as low ($, <$150K), medium ($$, $150K-$750K), or high ($$$, >$750K). These estimates define the bubble size in the graph in Appendix E and are noted for each service above and below. Three additional services emerge as a result, each with high strategic impact and lower estimated cost:

  1. Develop a self-service tool to guide service selection based on data classification, access

attributes, etc. (like Cornell’s “Finder tool”). ($)

  1. Provide secure DCC services that are functionally equivalent to OIT's existing virtual machine (VM) and other offerings. ($)
  2. Support faculty startup packages/semi-autonomous sub-clusters, delivering priority / immediate access to ‘owners’ while expanding the DCC and leveraging spare cycles. ($)

 

These twelve enumerated services reflect both needs of researchers and institutional priorities as conveyed by sponsors. They are recommended to advance to Phase 3, for the purpose of identifying appropriate ongoing (and as needed, bridge) funding to implement each service, ideally by FY25, either as pilots or production services.

 
  

 

3 NB: Eight lower-rated services were excluded from the Faculty Survey but were placed along the X-axis of the Appendix E bubble graph based on Poster Session ratings of those services, relative to other service ratings.

4 Sponsors rated all 29 consolidated services/projects, incorporating the eight lower/deferred priority items not presented in the faculty survey. This was in recognition that there could be certain services of high strategic value to the institution, but which would not necessarily be highly rated by individual faculty.

These twelve services recommended for Phase 3 form three broad service clusters, with the number beside each service corresponding to its overall priority as determined by the process detailed above.

 

Services organized into three clusters

On-the-Bubble Service Priorities: Two other services are worthy of further evaluation: both had reasonably high ratings by sponsors (4.0-4.25), but each received slightly lower faculty ratings and has an estimated annual cost that requires further financial validation:

    • Create Data Continuity Services that ensure data integrity and availability, including providing the storage associated with maintaining data continuity. ($$-$$$)
    • Create a single, central protected research network rather than the separate ones provided by OIT and DHTS. ($-$$)

Of note, a project to pilot moving a Basic Sciences unit to OIT’s network is being explored.

 

These two bulleted services, while not recommended as initial targets for Phase 3, deserve further study as to their approach, feasibility, and cost. Teams from ORI, OIT, Libraries and DHTS should more fully evaluate each, with a goal of developing detailed service proposals for consideration and / or pilot implementations in FY25.

 
  
 

 

School Based Priorities: Variation in priorities across scholarly domains motivates further discussion. Appendix G gives more detail on services with high domain-specific ratings but not promoted in the global process. One service highly rated by Basic Sciences/Nursing (create a single, central protected research network) already appears in the On-the-Bubble category above, with its strategy identified.

In addition, five other services rated highly within one or two domains, but not overall. These reflect collaborative opportunities with specific service partners (designated to the right of each service):

 

 

 

 

    1. Improve and clarify storage and computational options approved for regulated research, highly ranked by both Engineering and Social Sciences/Policy. ($)
    2. Add 4th data classification; where feasible, ease requirements on non-regulated sensitive data, a priority with Basic Sciences/Nursing. ($)

    with ORI, OIT and, Libraries:

    with OIT:

    Create an education cluster with CPU and GPU virtual machines to support course needs, among the highest priorities for Engineering. ($)

  1. e) Facilitate Cloud AND On-Prem options, including SOM researcher access OIT's services for non-clinical research where relevant, sought by Social Sciences/Policy. ($)

    These five lettered services are not recommended to Phase 3 but reflect partnering opportunities among sponsors and the deans. Ideally, such collaboration can lead to school-specific / bounded deployments in FY24 or FY25 that could be funded at the unit level, and with the potential of perhaps later rising to institutional-level services.

     
      
     

    w/ ORI,

    OIT &, DHTS:

    Improve web-based access to DCC resources, highly rated by Humanities/Arts. ($)

 

 

Of course, other services not called out above as collaborative opportunities can certainly be pursued, especially where the service has low estimated implementation costs or where creative approaches to implementation might be pursued. Consider, for example, two other services that were of particular interest to faculty in Natural Sciences/Environment (Create/optimize a special- purpose VM environment for graphical intensive work) and Engineering (Formalize/extend special purpose FastMPI cluster). In both cases the sponsors perceive opportunities to develop proposals with domain faculty to fund these services via foundation or federal agency grants. In the case of the Engineering FastMPI cluster, the graph in Appendix E shows two points, differenting its two cost approaches: one if funded institutionally and the other reflecting funding through grants.

This type of creative partnering could lead to implementation of considerably more than the twelve services initially prioritized and advanced to Phase 3. Of significance, this Research IT Needs process has already spurred creative implementation approaches for two pilots with Engineering. The FastMPI cluster referenced above is being funded primarily by Engineering, then opportunistically, an education cluster (School priority (c) above) will be created through the “trickle down” of some legacy Engineering MPI equipment, already operated by OIT. Each cluster’s usage model will follow the DCC model, where a primary queue provides priority access to the designated Pratt function(s), and lower priority queue(s) provide other researchers throughout Duke with ‘scavenger’ access to unused computational nodes.

 

Conclusion and Next Steps

The twelve services enumerated on page 3 now advance to Phase 3, due to their high likelihood of enhancing Research IT (and related) support across Duke. These services will form a coordinated and integrated research support program across the Office of Information Technology (OIT), Office of Research and Innovation (ORI), Duke University Libraries (DUL) and others, in conjunction with Schools. They will be pursued via the three Service Clusters articulated on page 4. Service partners have already launched initial work for five of the twelve priority services with low- cost estimates ($), even ahead of the essential and anticipated funding commitments in Phase 3.

Although implementing these services requires an incremental multimillion-dollar investment, the process through Phases 1 & 2 reinforces that even despite Duke’s extensive existing investment in research, new technological, regulatory, and competitive challenges demand appreciable, further investment for Duke to remain preeminent among research universities.

Phase 3 will aim to develop funding strategies for these services, in aggregate and individually. This will likely involve a combination of allocated funding, F&A/Indirect rate changes, services billed direct to grants, philanthropy or other approaches. Financial experts, along with sponsors and other leaders, will guide the Phase 3 work and develop financial proposals over the first quarter of FY24.

In parallel, two services identified on page 4 that do not advance to Phase 3 will be more fully studied by service providers / sponsors as potential targets for pilots in FY25. Further evaluation will refine their service definitions, cost estimates, and / or identify alternative approaches.

Finally, five further services designated as (a)-(e) on page 4 (and possibly others), reflect potential partnering opportunities among sponsors and individual Schools, at a pace and scale determined by the parties and based on localized priorities and resources at the service provider and School level. These services are estimated to be modest in cost (estimated <$150K), but because their value—at least initially—is believed to be more localized, they become opportunities to be funded by Schools.

The sponsors acknowledge and thank the many faculty and staff who have contributed to the process to date and are optimistic that after Phase 3 and the twelve initial service implementations, that some non-prioritized services above may become candidates to pursue in a future stage of this project.