1. Motivation

Over the past few decades, computational science and engineering has emerged as an important driver for all science and engineering disciplines. The US Department of Energy Office of Science asserts that: “Advances in the simulation of complex scientific and engineering systems provide an unparalleled opportunity for solving major problems that face the nation in the 21st Century.” (Scientific Discovery through Advanced Computing, 2000).

Computational science and engineering is an inherently collaborative and interdisciplinary endeavor. Computational researchers develop and use state-of-the art computing algorithms and technology to seek a deeper understanding of important problems spanning traditional science and engineering fields. A successful computational science and engineering education program must provide research support and training in computational skills and analytical problem solving methods. This knowledge is then synthesized to study basic and applied problems. Consequently, students require fundamental skills in mathematics, computer science, statistics, and a broad variety of areas in the social sciences, natural sciences, and engineering. Computational science and engineering does not replace or simply follow from experiments and empirical investigations; instead, theory, computation and experiment are complementary forms of scholarly inquiry and knowledge discovery. It is therefore crucial to the success of science and engineering research programs on this campus to promote and develop a comprehensive research profile that includes theory, computation, and experiment.

There are several strands of computational science and engineering already present at UC Merced. In fact, computational research occurs across all three schools as demonstrated by the list of faculty authoring this initiative. Data science and engineering is a new area within computational science and engineering which aims to analyze and interpret the enormous amounts of data produced by new, high-throughput technologies. Moreover, future computational faculty are key priorities for hiring in Management (business analytics), Applied Philosophy, and Life and Environmental Sciences (ecological theory/modeling). These future hires bring great potential to open new exciting avenues for on-campus collaboration with current computational scientists and engineers.

2. Establishing a new Centralized Research Unit

We propose here the formation of a new Centralized Research Unit (CRU) called the “UC Merced Center for Theory and Computation” (CTC). This CRU is a consortium of research groups and units already existing on campus. Going forward, this CRU will incorporate new research groups and units as they develop. The purpose of the CTC is to bring faculty and students together to address common research and training goals.

The CTC will seek to provide student support and training through undergraduate and graduate certificate programs in scientific computing. These certificate programs will supplement an existing undergraduate or graduate degree through additional courses that provide fundamental computational problem solving skills and experiences implementing and optimizing codes on high performance computing platforms. Additionally, the CTC will run a seminar series aimed at promoting computational science and engineering research on campus. Each faculty member of the CTC will contribute to this seminar series and distinguished colleagues from other universities and institutions will also be invited. In addition to providing students with exposure to computing research methodologies, this seminar series will promote collaboration among computational faculty across the disciplines.

Another way to leverage potential gains is through collaboration of the CTC with the national labs. There is a great interest from scientists at these labs to interact with faculty and students on campus. By providing this opportunity for lab scientists, we may facilitate new pipelines of research opportunities including student summer internships, collaborative research proposals, and access to state-of-the-art computing resources available at national labs.

The CTC may also partner with local CSUs that do not have sufficient on-site resources for training computationally-minded CSU students. This could serve as a recruiting strategy to bring well-trained graduate students into the computational sciences at UC Merced.

3. Resource requirements

Given the reach of this proposed CRU and its potential impact on the overall research stature of UC Merced, the resources required to launch the CTC are relatively modest. We discuss these resource requirements in detail below.

Research computing resources

To support the CTC, it is absolutely crucial to acquire and support campus research computing resources. These resources include on-site hardware, software, and administrative support staff. The CTC will develop innovative initiatives for sharing these research computing resources across all of the faculty in the CTC. At present, research computing resources are lacking in coordination and organization leading to unnecessary expenses and ineffective use. Most research computing on campus is done with individual faculty vying for space, power, and cooling resources within the limited space afforded by the server room in the Science & Engineering Building and in the absence of appropriate system administration staff. The CTC will organize and lead efforts to provide faculty a means to prioritize research computing needs for the campus and help to develop realistic plans for hardware and software upgrades as well as policies and procedures for shared research computing resources. This sharing strategy may be highly effective for purchasing software licenses and for testing scientific code on new hardware architectures, for example. For hardware, the CTC will need a few shared small- and medium-scale computing platforms for development and testing of codes, training of students, and performing scale studies required for transitioning to large-scale computing platforms. For large-scale computing needs, faculty will make use of national supercomputing centers such as XSEDE (https://www.xsede.org/).

The CTC should be supported by staff responsible for maintaining the computing hardware and software, ensuring stability in data storage and backup, and monitoring the security of the campus research computing systems. Initially, one full-time staff person will be required for resource support. As the CTC grows, additional staff may be necessary. This support staff can also provide bridge training to non-specialist researchers in using these computing resources, enabling researchers to obtain the necessary skills and initial results to prepare successful proposals for time on external large-scale computing platforms. A search is currently underway to hire a research computing staff person. The successful hiring of this position will go a long way toward filling in this missing resource.

While some may argue that all research computing can be performed off-campus at supercomputing centers or with cloud-computing resources, this is simply not correct. Computational science and engineering research encompasses faculty who are developing new simulation codes to solve complex science and engineering problems. Those developers require local computing resources to design, test, and validate programs prior to seeking supercomputing access. For example, XSEDE proposals stringently require preliminary data including verification of a working code, resource scaling studies, as well as clear identification of specific computing needs, e.g. memory, CPU, and/or communication intensive. Having medium-scale local computing resources also allows for a much faster turnaround for all research projects that require simulation or data processing, and may encourage researchers to extend the scope of their analyses leading to more rigorous computational testing of hypotheses. Moreover, local research computing resources are absolutely necessary for training of students in computational science and engineering. To deny computational science and engineering researchers on-site research computing resources is to deny the ability of these faculty members to conduct their research.

Full-time equivalents (FTEs)

We do not propose separate, formal request streams for new faculty FTEs to start and develop the CTC. Rather, the CTC will develop through support from existing research units and groups who already contribute to the CTC. For example, the entirety of the Applied Math Unit, and all of its future hires, will participate in the CTC. The Chemistry, Physics, Cognitive and Information Sciences, Quantitative Systems Biology, Electrical Engineering and Computer Science, and Mechanical Engineering and Applied Mechanics groups each have members who will participate in CTC and have future plans to hire faculty who will be potential CTC members.

Student certificate program in research computing

To develop the CTC into a student training program, we require the resources to offer classes, workshops, and seminars on topics that develop the common skillset of research computing.

Common space

The CTC space would ideally contain contiguous offices for CTC faculty, postdoctoral researchers, students, and visitors, a seminar room, and collaborative conference space for group meetings.

4. Comparison programs

UC Santa Barbara, UC Davis, and UC San Diego all have graduate programs in computational science and engineering. In addition, Stanford University has the Institute for Computational and Mathematical Engineering. There are additional programs throughout the nation that are comparable.

There are noteworthy differences in what we propose here from the examples mentioned above. One is a shared focus on undergraduate and graduate training. A recent NSF Mathematical and Physical Sciences report from the “Data-Enabled Science in the Mathematical and Physical Sciences” workshop on March 29-30, 2010 stated that these computational skills “must be percolated into lower levels of the curriculum, to train data proficient scientists in anticipation of a profound shift of research resources into data-enabled science in the future.” By providing undergraduates with training in computational science and engineering, we will become a recruiting ground for the aforementioned national programs. A strong academic program in this area will help attract well-prepared and motivated students to our campus.

The other significant difference is the close collaboration with the national labs. Through the development of a certificate program in computational science and engineering, lab scientists will have an immediate understanding and confidence in the abilities of our students when considering summer research opportunities, internships, postdoctoral research positions, etc. Moreover, by including lab scientists in the training process, we will create a natural means through which collaborations and working relationships will develop.