Duke University Research IT Needs Assessment Summary Report and Recommendations

Image

After feedback is received from University Leaders and Deans, it is anticipated that leaders from concomitant units (Research, IT, Libraries, etc.) will collaborate to develop proposals aimed at addressing these findings from the services perspective.

Introduction

Duke’s Information Technology Advisory Council (ITAC) in conjunction with staff support from the Office of Information Technology (OIT) has undertaken a comprehensive assessment of the research computing support and service requirements for campus researchers. Deans with responsibility over natural sciences, basic (life) sciences, social sciences, humanities, arts, and engineering have each selected groups of faculty to represent their own information technology (IT) needs, as well as those of their colleagues. These faculty members provided feedback on how to improve technology offerings and services to better support research needs, as well as the increasingly complex compliance and regulatory environment around the production, use, and preservation of research data. See Appendix A for a full list of the participating 39 faculty/research staff.

From February to August 2022, an OIT project team held listening sessions with seven groups of faculty and researchers in

• Natural Sciences
• Social Sciences
• Basic Sciences
• Engineering
• Humanities/Arts (Three Working Groups) Digital Environments, Digital Representation, Digital Pedagogy

After each listening session, a summary was prepared by the faculty participants (Appendix B) and shared with ITAC for discussion and elaboration on common needs (Appendix C). Two “poster” sessions were held in June and August to distill faculty feedback (Appendix D) into the top priorities and needs (Appendix E).

Through these appendices, the interested reader can be more fully immersed in the process as it progressed.

 

Findings and Concerns

All seven sessions began with a focus on technology concerns, and while the conversations highlighted improvements that could be made to technology-related services, each of the groups also noted the need for specialized support in the different disciplines and training, workshops and consulting for use of existing and new techniques or technologies (often referred to as "domain expertise"). As one faculty member noted, “A lot of things that look like computational problems are not, they should be done by experts in the field.”

Faculty representatives noted concerns about who is paying for what services and technologies—Big Duke or a department, or is it left up to the faculty member or a grant? In addition to costs and access to domain- specific support, faculty are also concerned about the ways in which organizational structures and policies may impact their conduct of research, such as processes associated with research administrative tasks such as completing data management plans or grant proposal development, submission and reviews (especially those involving sensitive data). This also encompassed concerns about the confusion or frustration that can be faced when navigating our complex ecosystem of inter-related, or sometimes siloed, offices and processes.

This document presents the summary observations of the working groups along with general recommendations to progress each observation (finding) in the categories of:

People - Identify skill sets and support staff investments needed to support research IT needs. Processes – Evaluate existing or implement new processes to align administrative tasks with research needs. This category also encompasses observations related to IT policies or organization structures. Technology - Prioritize the expansion of an existing or addition of a new technology services offering to help meet research needs.

Summary observations and recommendations below include links to the fully detailed findings (deeper insights from each group), which appear in Appendix F. Those details are an essential aspect of the process and integral to this shortened report but are shared via an appendix so as not to overwhelm or exhaust the reader.

Collectively, findings and recommendations reflect a need for new models and sustained support, not simply a one-time infusion of funds or point-in-time set of changes.

 

People Findings and Recommendations

Consistent across all the seven faculty working groups was the need for domain specific expertise. The OIT engagement team facilitating the research review had anticipated this unmet need from the humanities, arts and social sciences, but also noted the strong expression of need for research personnel support that was communicated by engineering, natural sciences and basic sciences. Most units praised their local IT support and general OIT services, but also communicated the need for specialized support personnel who could more fully engage the central IT structures, technical offerings, and policies, as well as offer training assistance. A few exemplar comments related to personnel appear below; a complete list appears in Appendix F (page 24).

• “The biggest need is for human infrastructure.” (Hannah Jacobs in ITAC discussion with Humanities/Arts)
• There is a need for more specialized/dedicated local support personnel who can quickly and efficiently address problems as they arise, without the need to bounce a ticket back and forth with others. Current ticketing process and search for expert help is arduous and inefficient. (Basic Sciences Faculty Working Group)
• People are important and some degree of domain knowledge is also really important—what physical chemists need, or what particle physicists need, or what genomicists need—they're not really the same thing. Compared to some of our peers we are under-resourced in that kind of domain computational expertise. The funding models of our peers are all over the map, from fully allocated to heavily charged out, but just in pure body count we have fewer of those people than many of our peers do right now. (ITAC discussion with Natural Sciences)
• An important thread is that Social Sciences (and others) have issues around involving novice graduate and undergraduate student researchers in projects. This is may not be an OIT issue per se, but it’s an important layer of our digital ecosystem that can get lost. (Paul Jaskot in ITAC discussion with Social Sciences)

Finding A: The focus of OIT’s “Research Computing” function as the prevailing embodiment of IT support for research established a robust and responsive infrastructure comprising networking, computation, and storage supporting general needs. However, it did not build a base level of personnel support for research endeavors that rely on data collection, analysis and storage or the application of technology to specific scientific domains. By leaving the human component to the lab, department, and school level, the result has been inconsistency in specialized research support as conveyed by all faculty working groups, even those with substantial extramural funding. See Appendix F (pages 24-27) for detailed insights reflecting all seven faculty working groups and in support of these recommendations.

Recommendation 1: In consultation with faculty, Duke should devise a service that builds and supports teams of domain-specific technical personnel who are well versed in both on-site and cloud options. The personnel should be equipped to assist with specialized needs in the different domains, including compute, storage, data collection and analysis, and new applications of technologies (e.g., applying machine learning, creating digital environments, and building and maintaining large, shared data resources). Evaluation of support structures should extend beyond department vs. central IT resources, also encompassing prospective engagement from areas like Libraries or other programs. The effort should start with pilots and expand with experience to take advantage of early lessons, include faculty feedback, and have central assistance in recruiting personnel.

Recommendation 2: Develop and catalog training resources to support an ongoing education program targeting faculty and graduate students but also serving undergraduates who participate in research endeavors. Such a program should apply established approaches (e.g. data analysis using common applications such as R and emerging machine learning) and should leverage the domain-specific technology support resources, as well as existing co-curricular programs such as via Libraries (e.g., CVDS), Center for Computational Thinking (CCT), OIT (e.g., Co-Lab Roots), and various curricular programs offered by academic departments, and others.

Personnel-related findings of interest to one or more groups, but not reflecting broad need, included:

• Centralized assistance for machine learning and deep learning applications like image analysis are needed (Natural Sciences)
• Include consultative support for grant preparation related to these Data Use Agreement requirements (Social Sciences)
• Re-evaluate basic technology support and research support levels and allocated resources in the humanities and arts and benchmark against programs at other leading universities. (Humanities/Arts)

 

Process and Organizational Findings and Recommendations

Duke University is a highly rated and well-respected research university, built upon the foundations of a liberal arts college and enriched by the incredible strength of the professional schools, including the Schools of Medicine (SoM) and Nursing and notable others, as well as the deeply technical disciplines represented by Engineering, Natural Sciences and Social Sciences. However, the impact of Duke multi-disciplinary academic and research potential is severely blunted by different or siloed organizational structures, inconsistent or competing strategic directions, and conflicting policies that restrict or make difficult the collaboration that undergirds all scientific discovery. Throughout all groups we heard these concerns noted, but most vocally from Basic Sciences and Natural Sciences. Sample comments related to Process and Organization appear below; a full list appears in Appendix F (page 27).

• Seams are beginning to show in the overall research IT service model. There is a growing lack of understanding between OIT and SoM IT providers as to what services are provided by whom and how to acquire them. (Natural Sciences Faculty Working Group)
• Better mechanisms are needed for communicating IT users’ needs/requirements to the policy makers and financial managers. ITAC is an excellent mechanism for this on the University side, but there is need for a comparable avenue in SoM. (Basic Sciences Faculty Working Group)
• We need a better connection of local support personnel (with domain knowledge) to the larger IT support system, be it DHTS or OIT. Today we have a system where we file online tickets and although everybody has the best intention, sometimes it's hard to find the person who's qualified and has the security clearance to actually implement the solution and oftentimes it comes back to us. It's not super-efficient when you have to interact with 6 people, and they are doing their best with everybody trying to locate the right person. I would love a simple system where we know directly who knows enough to solve the problem and maybe could have the security clearance to fix my problem. (ITAC Discussion with Basic Sciences)

Finding B: Separate IT infrastructures supporting research have operated for more than two decades (OIT-Campus Schools, DHTS-Schools of Medicine and Nursing), along with distinct lab and departmental IT personnel. Despite each organization striving to deliver great service, and while OIT and DHTS coordinate on many items, the impact for faculty collaborating across these complex infrastructures for research has become increasingly problematic. For example, decisions of SoM research policy and support are well understood to be the purview of the School, however their impact can reverberate across campus in terms of research administration and policies as well as directly impact faculty in campus schools such as Pratt and Trinity. See Appendix F (pages 27-30), for detailed insights which were most vocally expressed by Natural Sciences and Basic Sciences, but also echoed by other faculty working groups.

Recommendation 3: Conduct an external review with peer expert leaders (from academic medical centers that are tightly coupled with their universities) to objectively assess costs/benefits of dual and decentralized research infrastructures (OIT/DHTS/Schools/Institutes); present for review to University leadership including the Provost, SoM Dean and other University and DUHS leaders. Regardless of the outcome of the review, seek to discontinue or drastically limit the practice of SoM policies and technologies being imposed on the campus without input from faculty.

Finding C: Disparate and sometimes conflicting compliance and security measures (Campus and SoM) and a seeming lack of gradations of risk for various data types and uses can inhibit intra- and inter- institutional research collaboration. While researchers acknowledge and accept the necessity to comply with requirements to protect sensitive and restricted data, they desire approaches that better define options and balance the risk of actual data loss with the perceived adverse impact to Duke’s research endeavor itself. See Appendix F (pages 31-33), as conveyed to varying degrees by the majority of faculty working groups.

Recommendation 4: In collaboration with faculty representatives, the VP for Research and Innovation together with the DU CIO should launch an evaluation of current policy, security and compliance requirements and processes for research toward a holistic risk-based institutional approach. The goal should be to look for minor adaptations to existing policies or seek new solutions (if warranted) in cases where the adaptations better balance research practices / needs with risk, but without compromising regulatory requirements or the protection of the institution against sensitive data loss.

The effort should also reflect the need for nuanced policies or implementations for social science and campus researchers working with sensitive data that are not predicated or based on similar DUHS, School of Medicine, or School of Nursing clinical research policies or implementations, and to take care not to conflate clinical studies with other research endeavors. Ongoing research-related governance bodies should include sufficient faculty participation to incorporate representative feedback across the diversity of Duke’s research enterprise and work towards increased transparency about the processes and decision making.

Other Process and Organizational findings that had limited rather than universal support included:

• Communication and coordination are needed across the many oversight bodies as the process for the researcher has become complex, time-consuming, and a hindrance to research. (Social Sciences)
• Greater representation is needed from social scientists at the decision-making table, particularly regarding policies that impact Social Sciences research. For example, the updated Data Policy guidelines. (Social Sciences)
• Create support structures for team teaching in the humanities that may become a means of establishing initial ‘connective tissue’ among faculty and with IT (Humanities/Arts)

 

Technology Findings and Recommendations

While faculty across all working groups were aware of and praised the Duke Compute Cluster service and associated personnel, there was no shortage of concern expressed that the current offerings are too limited in terms of flexibility and scale of the compute offerings, the flexibility and cost of storage solutions over the lifecycle of research projects, and the technical guidance available to help discern and implement the best fit solution among local (on-premises) options and cloud approaches. These concerns were most pronounced— to a person—throughout Engineering, Natural Sciences and Basic Sciences, but were echoed by individuals on the Social Sciences and Humanities/Arts working groups. Sample feedback regarding Technology themed findings appear below; for a complete list see Appendix F, page 33.

• Many researchers are having to go outside Duke to peer institutions and national labs to use certain clusters; it would be a boon to the University and researchers alike to have the means to run small to medium scale jobs with a home cluster. (Natural Sciences Faculty Working Group)
• Several years ago, the Provost agreed to commit to covering 80% of all the computing needs of the faculty, recognizing the 20% of the most computationally intensive users would need to utilize grants or other approaches to meet those needs. Recognizing those needs may have changed over time, we should reassess whether the level of “entitlement” computing being provided by OIT still meets that 80% threshold. (ITAC discussion with Natural Sciences)
• There’s an unmet need for mid-scale computing resources that would provide more than the Duke Compute Cluster but less than provided by a national lab. This includes fast interconnection and a diversity of hardware. (Engineering Faculty Working Group)
• NIH funded data storage retention requirements need to be recognized by both DHS and DU as a critical component of infrastructure. (Basic Sciences Faculty Working Group)

Finding D: The Duke Compute Cluster (DCC) and RAPID VM services, while quite valuable and cost effective, are not as expansive as many faculty require. For some, it is difficult or less attractive to use for certain teaching and specialized research tasks. On a continuing basis, and as our needs and technology evolve, OIT should work with ITAC and academic leadership to evaluate and propose changes or additions to research computing services, and to enhance the guidance OIT offers to its users. See Appendix F (pages 33-36) as conveyed to varying degrees by all faculty working groups.

Recommendation 5: Evaluate the cost and funding approach to extend the capability of Research Computing by expanded HPC/GPU services so that they are better sized to meet demand and are easier to use; review with ITAC and academic leadership to determine viable approaches that can improve support and are financially feasible. Establish a more routine review cycle with ITAC faculty (or other faculty advisory bodies) to better evolve services with need.

Recommendation 6: Establish one or more Faculty Working Groups to help guide OIT in tuning Research Computing offerings to better support data collection, statistical analysis, and machine learning, as well as common programming and graphics-intensive needs. New approaches need to be considered for the extremely data intensive needs of instrumentation, along with the financial models that can support them (funded via grants, indirect cost recover overhead, subvention, or otherwise).

Finding E: Insufficient awareness exists about computational options and there is uncertainty of when to use which as many options are often presented and left to the faculty to choose. This lack of awareness, coupled with complexity in navigating the options, stymies researcher ability to easily access the resources they require. See Appendix F (pages 36-37), as conveyed to varying degrees by all faculty working groups.

Recommendation 7: Introduce guidance and simplify navigation of technical solutions for faculty and students by consolidating technical offerings to more broadly support research and academic use cases or needs (including labs with substantial data generation or acquisition from specialized instruments). Solutions should: encompass the full breadth of services available, including local and cloud options; be established as an evolving service; and be developed in consultation with faculty to ‘field test’ the efficacy of the guidance and navigation across the range of Duke’s research domains.

Finding F: The needs for research data storage have increased exponentially; current storage services address the need for active, near-term research storage but are not easy enough to use and share and do not incorporate solutions that are affordable for long-term and archival storage. Moreover, current options do not reasonably accommodate NIH and NSF ‘unfunded mandates’ for long-term data retention in ways that are either scalable or cost effective. See Appendix F (pages 37-39), as conveyed to varying degrees by all faculty working group (Recommendations 8 & 9) or as may represent an opportunity broadly for Duke but introduced most explicitly by Social Sciences (Recommendation 10).

Recommendation 8: Implement new long-term storage options for a variety of use-cases and ensure solutions are available to not only campus but also School of Medicine and Nursing researchers; establish funding models with leadership and faculty to ensure sustainability.

Recommendation 9: Ensure the new long-term storage options along with existing and future storage solutions are approached as an eco-system, recognize the differences in storage needs across the research lifecycle and automating to the greatest degree the movement of the data across solutions to reduce the burden on faculty and risks to Duke. Make these options available to both campus and SoM users with attention to differences in datatypes and data storage needs in the science versus humanities versus social sciences.

Recommendation 10: Evaluate the need and develop a proposal with estimated cost to establish a new service for researchers that would manage/license/administer shared access to expensive data resources/sets; carry out this work in collaboration with the prior Presidential Scholar who recently studied this need in greater detail on

of the Provost and President and in conjunction with the Strategy Team 2030 implementation team.

Other Technology findings that were less universal included:

• Provide additional physical computer labs (like Link Classroom 6) so students are not faced with using small laptop screens to access VMs for graphics-intense software like GIS. (Humanities/Arts)
• Provide faster interconnects for the DCC (needed to support certain kinds of research) (Engineering)
• A need for “bare metal” hardware environments, beyond what the DCC provides (Engineering)
• Better digital publishing services, especially extra-mural options w/ visualization + archive options; related, digital scholarship may require specialized tools and platforms for archiving. (Humanities/Arts)
• Establish clear guidance/direction for existing services for data collection, storage, and compute services that are available to researchers at no or low cost; clearly classify their appropriateness for use based on the sensitivity of the data or research being performed (Social Sciences, Basic Sciences)

 

Conclusion and Next Steps

To improve research IT support in an environment of ever-increasing requirements, complexity, and interdependence, will require a multilayered approach that focuses on incremental improvements to offer near-term relief while also moving forward on systemic issues. Broad input from Duke’s faculty and researchers indicate the need to:

• Simplify services and reduce the amounts of time, effort, and money that researchers need to spend managing Information Technology and navigating among different offices or policies.
• Provide more layered services that match evolving needs of researchers based on acceleration of regulatory and technology requirements. (i.e., data management services vs. storage)
• Offer domain specific services, protocols, and people to better match evolving needs in different research domains. Universal across all sessions was a clear cry for more domain expertise from computational support staff.

Faculty urged repeatedly and vocally throughout the process that at least four areas of overall commitment are needed related to governance, financials, or service span in order to protect Duke from repeating the history that led to the current situation (a significant gap between research needs and support levels):

1. A faculty-centric oversight or governance body (or bodies) to help monitor and advise when and how services need to evolve, policies need to be established or revisited, etc.
2. A sustained budget commitment reflecting research support as a continuing and evolving need, requiring not just the one-time funds but also ongoing financial commitments and flexible financial models to address the changing needs and environmental factors (and research constraints) over time.
3. Recognition that future and ongoing services in support of research should avoid bright dividing

lines or “firewalls” between Schools of Medicine/Nursing researchers and the rest of campus.

4. Clarity on overall process ownership of the organization(s) with responsibility, authority and accountability to sustain and evolve research support needs such as those characterized herein; when support or processes span multiple organizations, a commitment to a One Duke approach to solving.

An essential next step will be to share these findings with the greater Duke community to ensure feedback is consistent, validate that the recommendations (process oriented) are embraced by university and school leadership (Provost, EVP, Deans, and others), and identify and align partners who will help to implement recommendations, notably expected to include: VPRI, University Librarian, CIO, others, and the faculty bodies who will help to guide the implementation of recommendations (ITAC and others).

Following the validation process and report issuance, specific further steps are not proposed as part of this report. Rather, specific service and support proposals responsive to the report’s findings and recommendations are expected to be developed over the coming months collaboratively among the CIO, VPRI and University Librarian.

The Information Technology Advisory Council, the Office of Information Technology, and the Office of the Vice President for Research and Innovation sincerely thank the many individuals who participated in the assessment project and contributed to this report.

Appendix A: Working Group Faculty Participants (Identified through Deans)

Natural Sciences (Identified by Mohammed Noor)

David Beratan, Chemistry

Eric Laber, Statistical Science, Biostatistics & Bioinformatics, and Global Health Jianfeng Lu, Mathematics, Chemistry, and Physics

Dan Scolnic, Physics

Jenny Tung, Evolutionary Anthropology, Biology and DUPRI

Greg Wray, Biology, Evolutionary Anthropology, and Biostatistics & Bioinformatics

 

Social Sciences (Identified by Rachel Kranton and Don Taylor)

Kate Bundorf, Sanford School and Margolis Center for Health Policy

Sunshine Hillygus, Political Science and DISM (Duke Institute for Survey Methods) Jim Moody, Sociology and DNAC (Duke Network Analysis Center)

Daniel Xu, Economics and Triangle Census Data Research Center Chris Bail, Sociology and Duke Polarization Lab

 

Basic Sciences (Identified by Colin Duckett)

Lisa Cameron, Biology and Director, Light Microscopy Core

Alexandra Badea, Radiology, Neurology, Biomedical Engineering, BIAC Greg Crawford, Pediatrics / Molecular Genetics and Microbiology Lindsey Glickfeld, Neurosciences

David Herzfeld, Neurobiology

Danny Lew - Pharmacology and Cancer Biology

 

Engineering (Identified by Jerry Lynch and Claudia Gunsch)

Volker Blum, Mechanical Engineering and Materials Science Cate Brinson, Mechanical Engineering and Materials Science Jessilyn Dunn, Biomedical Engineering

Helen Li, Electrical and Computer Engineering Miroslav Pajic, Electrical and Computer Engineering Henry Pfister, Electrical and Computer Engineering

Amanda Randles, Biomedical Engineering, Computer Science, Duke Cancer Institute

 

Humanities/Arts (Identified by William Johnson, Paul Jaskot, John Brown and Victoria Szabo)

Digital Pedagogy:

Astrid Giugni, English Thavolia Glymph, History

Hannah Jacobs, Art, Art History and Visual Studies (research/teaching staff) Laura Lieber, Religious Studies

Clare Woods, Classical Studies