Subject: EDUC - more Resent-Date: Wed, 01 Dec 1999 13:28:44 -0500 Resent-From: Geoffrey Fox Resent-To: p_gcf@npac.syr.edu Date: Wed, 1 Dec 1999 10:37:23 -0600 (CST) From: Joe Thompson To: gcf@npac.syr.edu CC: joe@erc.msstate.edu And this is an earlier White Paper I did last year for possible SURA participation in the DOE SSI effort. No action pending on this - just FYI and your use as needed. ----- Begin Included Message ----- >From joe@ERC.MsState.Edu Mon Oct 19 14:01:29 1998 From: Joe Thompson Date: Mon, 19 Oct 1998 14:01:26 -0500 (CDT) To: watson@jlab.org Subject: EDUC Cc: joe@ERC.MsState.Edu X-Sun-Charset: US-ASCII Content-Length: 34097 Chip, I worked this up right after the DC meeting on SSP, and sent it in to SURA. Joe ============================================================================ A WHITE PAPER for the DoE Scientific Simulation Plan SURA Region-wide Graduate Program in Computational Science & Engineering A Multi-Disciplinary, Multi-University Program with DoE Labs INTRODUCTION The emergence of computational simulation, enabled by high performance computing (HPC) and networking, as the third paradigm of scientific investigation has created the need for graduates trained across the disciplines of mathematics, computer science, computer engineering, and the various disciplines of engineering and the sciences. Via the networking that is inherent in this new paradigm, these graduates must necessarily be trained in teamwork and collaborative endeavor. A urgent need has thus been created for new curricula in computational science and engineering to meet the Nation's need for scientists and engineers with the broad understanding necessary to develop and apply these new investigative tools to scientific analysis and engineering design. This is especially true of the DoE Scientific Simulation Plan. The potential for computational simulation has been amply documented in recent years, and efforts toward convergence to a cross-disciplinary structure that can enable training and graduate education are emerging, witness the theme of the April-June 1997 issue of IEEE Computational Science & Engineering - and, in fact, the existence of this journal itself. The new computational simulation paradigm has created the need for cross-disciplinary education of graduate and undergraduate students to produce scientists and engineers with broad viewpoints and backgrounds to integrate the interaction of the problem domain with engineering design and computational methods and tools, including computer science, computing technology, and mathematics. It is vitally important for scientists and engineers to understand computational methods, technologies, and tools and to be prepared to apply these tools in the practice of particular disciplines. This new paradigm of scientific investigation requires innovative new graduate programs providing cross-training in science/engineering and in computer and computational science. This necessitates something of a change in university culture, since these new graduate programs must be inherently multi-disciplinary and, since various levels of expertise and experience exist at different universities, such new programs can best be mounted through multi-university effort in which the student spends time at different institutions and in which graduate courses are shared among institutions via the web. It is also essential that time in practice, such as at DoE labs, be included. The Southeast has emerged as a leading region in HPC in the US, and the region of the SURA universities now has 23% of the total unclassified HPC power in the country, and 43% exclusive of New Mexico and California. The universities of SURA are thus especially well-positioned both geographically and through ready resources and expertise to join with DoE to establish and operate a multi-disciplinary and multi-university graduate program in Computational Science & Engineering (CS&E). This multi-university graduate program will then serve as a major source of future personnel for the DoE labs in SSP, and the close association of universities in CS&E thus forged will provide a continuing source of expertise for DoE in SSP. In this program, graduate students will move among the SURA universities, spending time at more than a single institution during the course of the degree study. And students will take courses from the various SURA universities over the web, even while at the home institution. The SURA universities will compile a dynamic and expanding list of relevant available courses for this CS&E graduate program for which graduate credit will be freely transferred among the universities. The student will also spend time at DoE labs in the region and elsewhere during the research phase of the study. This graduate program will be guided my a board composed of representatives from the participating universites and DoE labs, under the auspices of SURA. A feature of the program will be premium fellowships to attract US citizens into the program. The program will also be designed to provide opportunities for professionals at DoE labs for graduate work in CS&E. Since the SURA region encompasses a large portion of the Nation's minority population, this regional graduate program will also serve to significantly increase the opportunities for minorities in the emerging field of CS&E. Finally, this multi-university graduate program in CS&E in the Southeast will serve to strongly influence undergraduate science and engineering curricula through the addition of neded CS&E components. PROPOSED CS&E GRADUATE PROGRAM Rationale The potential impact of computational simulation on scientific analysis and engineering design is revolutionary in both the extension of the scope of scientific investigation and in the reduction of the design cycle and the extension of the range of possible design considerations. But this impact cannot be achieved until graduates are available with the cross-training inherent in the concept of CS&E. And such training is beyond what can reasonably be included in a disciplinary BS program. But the present large difference between graduate assistant stipends and starting salaries in industry simply precludes any significant increase in the number of US students going on to graduate study. Consequently, graduate programs in CS&E must be instituted that both meet the cross-disciplinary requirement and surmount the reluctance of US students to continue into graduate study. And, since the impetus is to meet the applications need for strong capability in CS&E, there must necessarily be a strong federal lab and industrial involvement in the design and operation of such a graduate program. Currently, that inducement to graduate study, that federal lab and industrial component, and some of that cross-disciplinaty orientation are all missing from most available graduate programs related to CS&E. And most programs are only "related" certificate programs: there is the need now for specific graduate programs in CS&E, but still with discipline-specific applications. Finally, there is the clearly documented need to expand the student population into more diversity in order to meet the workforce demands of the next century. This graduate program in Computational Science & Engineering, proposed here as an extension and significant enhancement of an existing relevant programs at SURA universities, will address all four of these concerns: by joining with DoE labs to combine partial salary with a graduate stipend, by providing the option of theses and dissertations done on-site in DoE labs, by carrying the cross-disciplinary concept forward into multi-university operation, and by capitalizing on an established relationship with HBCUs/MIs in the Southeast. This graduate program will foster a close interchange of technology and experience as students from the DoE labs spend time at the SURA universities involved, and as students enrolling directly in the program spend time at the DoE labs. Web-based courses delivered remotely will be a growing feature of this multi-university graduate program, and this development will also serve to enhance opportunities for professional study in CS&E at the DoE labs. Multi-Disciplinary Research Theme and Major Research Efforts A computational simulation system requires geometrical representation, numerical solution, scalable parallel programming, and scientific visualization operating in a coordinated environment making efficient and effective use of parallel computing platforms friendly to the user. The SURA universities strategically address all these elements in such combination and with an applications focus through various established research programs that will feed to this graduate program. Operation This graduate program for cross training in engineering and the sciences with computational science & engineering will have three major component elements: (1) An inderdisciplinary graduate program in Computational Science & Engineering entered by students with undergraduate degrees in engineering or the sciences, (2) A post-doc program in Computational Science & Engineering for graduates with advanced degrees in engineering or the sciences, and (3) An undergraduate program in Computational Science & Engineering designed to attract students into the graduate program. Pervading all three components will be specific emphasis on enhancing the opportunities for minorities in Computational Science & Engineering, particularly through specific efforts to attract minorities into the program and to foster progress of minorities in toward success in the program. These three components will operate as follows: Existing PhD programs in Computational Science & Engineering at SURA universities will be coordinated, enhanced, and expanded into a region-wide program that is not only cross-disciplinary across mathematics, computer science, engineering, and the sciences, but that also is multi-university across the participating SURA universities with related programs, and that is directly opportunistic to diversity through involvement of HBCUs/MIs in the SURA region. Since MS and PhD degrees in Computational Engineering already exist at MSU, it will not be necessary to petition for the establishment of new degree programs before the program can begin operation, since degrees can initially be granted by MSU, transferring courses from other participating SURA universities. Formal degree programs will be initiated at other SURA universities as circumstances permit. The structure of this graduate program will be as follows: (1). Administration will be through a Board of Directors composed of representatives from the participating SURA universities and DoE labs, under the auspices of SURA. The individual universities will necessarily retain authority over their separate degree programs, but will collaborate through freely transferring and accepting credits for relevant graduate courses, and through facilitation exchange of students. (2). The program of coursework and the composition of the student's graduate committee will come from each of computer science, mathematics, and a specific discipline of engineering or science. (3). The student will choose a specific discipline of engineering or science for emphasis in the dissertation. (4). There will be the option of conducting the dissertation research at a participating DoE lab, under joint direction of MSU and the lab. In any case, the student will spend at least a summer, or equivalent period, on-site at a DoE lab during the course of graduate study. (5). Web-based graduate courses will be operated as an integral part of this graduate program. A specific option to be considered with the DoE labs will be the following: A BS graduate is hired by a DoE lab, but is placed in the MS program for one year before reporting for work. Salary during this period is reduced, but the usual MS graduate assistantship is available as well. The student spends three semesters at participating SURA universities. The thesis is completed at the DoE lab, and the degree is awarded after going to work. There may be some agreement as to remaining with the DoE lab for some period. This approach would cut through the present problem of the BS student being attracted away from graduate study by the starting salary - since the combination of the graduate assistantship and the reduced salary from the DoE lab should approach the full salary. A post-doc program will also be included, designed for graduates with PhDs in engineering or the sciences, utilizing appropriate courses from this CS&E graduate program in connection with research projects. A primary initial source for these post-docs will, of course, be the DoE labs, but recruitment will proceed on a broader scale. The summer undergraduate program will be modeled on the NSF Research Experiences for Undergraduates (REU) program, but with specific design to stimulate interest in CS&E in undergraduate students. In addition, undergraduate research programs will be continually operated at the participating SURA universities during the academic year, especially in connection with DoE-sponsored research efforts. Thiese undergraduate research programs will have the express purpose of attracting students into the CS&E graduate grogram. Evaluation This graduate program, and its attendant post-doc and undergraduate research components, will be specifically evaluated in terms of the number of students directly involved, the number of students supplementing other graduate degrees with courses from this program, and the placement of such students after graduation. All students - graduate, post-doc, and undergraduate - participating in the program in any way will be tracked. PARTICIPATING UNIVERSITIES ... COURSE LIST ... ------------------------------------------------------------------------------ PRECEDENT The Computational Engineering Graduate Program at Mississippi State http://www.erc.msstate.edu/education/compu_eng/index.html In response to this national need, Mississipp State University was selected by the National Science Foundation in 1990 to establish an Engineering Research Center for Computational Field Simulation. The Center was founded on two established research programs at Mississippi State -- computational fluid dynamics and microelectronics design -- which had come together under a pilot DARPA project. As a part of the support for the NSF award, the Mississippi Institutions for Higher Learning granted permission to the College of Engineering of MSU to establish a new graduate program to award the MS and PhD degrees in Computational Engineering. This is an interdisciplinary graduate program across engineering, computer science, and mathematics - housed in the ERC. Entry into this graduate program can be with a BS degree in any physical or biological science, or in engineering or mathematics. Computational engineering is the application of computational methods and high performance computing to solve large-scale complex scientific problems. It addresses problems that cannot be solved easily by analytical means or physical modelling, that have precise mathematical statements, that require knowledge of the discipline, and that are of significant scope. Examples of such problems are: analyzing the flow around a complex airfoil (plane) to determine stress under certain conditions, analyzing the behaviors of electromagnetic fields, analyzing the movement of pollution through ground water aquifers, or analyzing biological flows and processes. In fact, solving such problems has been identified by the U.S. Office of Science and Technology as some of the Nation's "Grand Challenges". This graduate program in Computational Engineering at Mississippi State incorporates instruction and experience in geometrical representation, numerical solutions, and scientific visualization operating in a coordinated environment making efficient and effective use of parallel computing platforms. The graduate program is designed so that students combine courses from the participating disciplines and from such areas as scientific visualization to focus on the computing, mathematics, methodology, tools, and application of computationally oriented analysis, design, and problem solving in science and engineering. Students are exposed to state-of-the-art numerical mathematics, high-performance computer architectures, software development tools for parallel and vector computers, field physics and grid generation methodologies, algorithm analysis and design, scientific visualization, and the application of computational techniques to at least one scientific or engineering area, such as fluid dynamics, electromagnetics, heat transfer, or structures. Since its inception in 1991, this graduate program in Computational Engineering at Mississippi State has produced __ MS graduates and __ PhD graduates. Of these, __ have has BS degrees in engineering, and __ have had BS degrees from other fields. __ of these graduates have been women. Graduates of the program are uniquely qualified and heavily recruited by major scientific and engineering laboratories in government and industry as specialists in computational analysis and design and as high performance computing specialists. As a part of the Programming Environment & Training (PET) effort at the CEWES MSRC in the DoD HPCMP, Mississippi State has initiated a variant of its existing PhD program in Computational Engineering whereby a student spends a semester each at NCSA (Illinois) and at CRPC (Rice), transferring credits from these institutions to MSU. In addition, the student conducts the dissertation research at CEWES MSRC under the direction of a committee formed and operated by MSU, but including ex-officio members from NCSA, CRPC, and CEWES MSRC. This program was initiated with one student in Summer 97. The PET effort at the CEWES MSRC, under the direction of the ERC, has initiated a major project in web-based remote training from Syracuse University (PET team member) to Jackson State University (PET team member) and to CEWES MSRC. Full semester undergraduate courses called were offered at Jackson State (for credit at JSU) during the Fall 97 and Spring 98 semesters, conducted remotely from Syracuse over the network. A graduate course is now being conducted. In addition to this graduate program, undergraduate science and engineering students wishing to so prepare themselves are offered a minor in Computational Engineering consisting of four courses. Two courses concentrate on computational field simulation and two courses focus on computing. Students completing these four courses will be awarded a minor in Computational Engineering. Virginia/ICASE/LaRC Program in High Performance Computing and Communication http://www.cs.odu.edu/~hpcc/ The purpose of the VILaP-HPCC is two-fold: to enhance NASA's ability to conduct computationally-intensive aspects of its research mission, and to educate applied computer scientists capable of lifelong contributions to "Grand Challenge" problems. Its programs are designed to attract graduate students of the highest quality with appetites for interdisciplinary work to the computer and applied science departments of participating Virginia universities and to NASA Langley, and to involve them at the thesis level in computational issues of long-term interest to NASA. The breadth of HPCC and the magnitude of resources required to provide a state of the art curriculum and research environment are such that, at the uppermost levels, a pooled effort leverages university and federal resources more effectively than duplicated efforts in achieving national impact. The term "High Performance Computing and Communication" has been appropriated for a variety of purposes since its coining in the federal initiative of the same name, ranging from computer science and computer engineering on one hand to computational science and computational engineering on the other. The former disciplines emphasize the development of software and hardware tools; the latter emphasize simulation of the physical world via discrete computational models. The development of tools (including libraries, environments, protocols, and devices) is often uninformed by the requirements of specific modeling communities, leading to tools that are not useful. Conversely, many modelers seek raw resources (megabytes and megaflops) without appreciation of how they are supplied, and hence compute and communicate inefficiently. The niche targeted by this program spans these two extremes, both philosophically and in implementation. It can be characterized as "computer science in the service of NASA applications."Because the research frontier of each computational application at NASA is highly specialized, few, if any, individuals will comprehensively integrate all disciplinary inputs, but some disciplinary bridging should be expected of everyone being trained in HPCC today. The Center's program will be attractive to students with backgrounds from all along the continuum described above who are interested in acquiring depth in a specialty but not at the expense of breadth. Significant computer science coursework will be required, but the topics of thesis research will be identified jointly by each student and a mentor from the computational applications community. The main research contributions are expected to be algorithms and productivity-enhancing environments for large-scale distributed parallel computation, scientific visualization, and scientific data base query. Unlike most university-based HPCC programs, this program is based at a major federal research laboratory engaged in several of the federally designated "Grand Challenge" projects. Mentoring from full-time NASA scientists and engineers will give graduate students direct contact with real-world problems, computer codes, and computer hardware. Being exposed to problems as they first arise in practice gives participating students a headstart over those who first hear about them at conferences after they have begun to be addressed. Such exposure also provides motivation that enlivens the otherwise abstract pursuit of a more optimal data structure or technique, and a "hard" test that validates incremental successes. (It is also acknowledged that there is a danger in getting absorbed too early in one's research career in a problem that is too narrow, and each student's faculty committee will be be composed carefully to ensure that the student's time is driven by educational opportunities more than milestones of the project in which the student is involved.) Unlike most joint university-NASA educational programs, this program involves student and faculty participants with an expressly computer science orientation. Most researchers are computer users; this program seeks fellows who will also contribute to the algorithmic and software infrastructure in a way that transcends a particular application, but will usually be motivated by a particular application in which its value is concretely demonstrated. Fortunately, it is now common for a scientist or engineer to have one, two, or even three semester-length courses in numerical analysis. However, it is rare for a scientist or engineer to have semester-length courses in nonnumerical algorithms, computational complexity, computer architecture, compilers, operating systems, communication networks, modern programming languages, modern computational paradigms (such as object orientation or functional programming), data bases, expert systems, and visualization. Insufficient exposure to the state of the art in these areas may block progress in complex problems no matter how deep the physical insight or brilliant the modeling creativity of a user. The Program is operated by an Administrative Director and a Council drawn from the faculties of the participating institutions and from NASA Langley. The Chair of each participating department or institute appoints one currently active faculty member to the Council. An equal number of NASA researchers sit on the Council, as appointed by the Langley Chief Scientist. An appointee of the Director of ICASE is also a voting member of the Council. The plan of study must satisfy three basic requirements: (1) it must ensure that the student stays on the track of an acceptable doctoral program in computer science or computational science at the host institution, (2) it must include within the first two years six credit hours (one semester-length course each) in two "core" HPCC areas, and (3) it must include within the first two years at least six credit hours (two additional semester-length courses) outside of the student's background area, chosen to give the student a basic foundation and a working vocabulary in both computer science and in an application area in which high performance computing is required. The two core courses, required for all participants in the program are: (1) high performance computer architecture, and (2) computational linear algebra. An example of the external courses would be two courses in fluid mechanics for a student preparing for computational aeronautics or atmospheric modeling. Specific approval of the external courses is delegated to the student's degree committee. ------------------------------------------------------------------------------- APPENDIX - Relevant Resources and Expertise in Participating SURA Universities Mississippi State University Mississippi has now moved into third position among all the states in terms of total unclassified HPC facilities within the state (two of the four DoD Major Shared Resource Centers [MSRCs] are located in Mississippi) and is the top EPSCoR state in this respect. The NSF Engineering Research Center (ERC) for Computational Field Simulation at Mississippi State University is still the only one of the NSF ERCs having its focus directly on high performance computing. Mississippi is also a national center of telecommunications, with WorldCom having its corporate headquarters in the state. Mississippi State University is uniquely positioned to mount a major cross-disciplinary and cross-university graduate program in Computational Science & Engineering because of (1). its NSF Engineering Research Center (ERC) for Computational Field Simulation, now in its eighth year, (2). the leadership position of the ERC in the Programming Environment & Training (PET) support program of the DoD Major Shared Resource Centers (MSRCs), now in its third year, and (3). the successful MS and PhD programs in Computational Engineering operated by the ERC, now with __ graduates and __ current students. 1. Composition & Scope of the ERC Initially funded by NSF in 1990, the ERC for Computational Field Simulation at Mississippi State is a multi-disciplinary academic research center - funded at approximately $9M annually by NSF, DoD, NASA, DoE, and industry - conducting a coordinated research program according to a strategic plan to advance US capability in the use of computational simulation in engineering analysis and design, as well as in scientific research in general. This Center focuses on all elements involved in the computational simulation of physical field phenomena: physical processes occuring over space and time, i.e. governed by partial differential equations - computationally intense simulations requiring access and efficient utilization of HPC facilities at the highest level, and requiring distributed graphics at the highest level for effective collaboration with other centers of effort. As an NSF ERC, the Center for Computational Field Simulation at Mississippi State has the mission of interacting with industry and federal labs in research of importance to economic competitiveness and national security. The specific mission of this ERC is to develop high-level capability for computational field simulation of physical problems for application in analysis and design. That mission is approached through research thrusts in five areas: grid generation, solution algorithms, scientific visualization, system software, and computer architecture - mounting an integrated research program in both software and hardware. This Center necessarily incorporates engineers, physicists, computer scientists, and mathematicians in cross-disciplinary research in geometrical representation, numerical solutions, and scientific visualization - together with the underlying parallel computing environments and mathematical foundations. Although the Center's historical concentration has been in computational fluid dynamics, its strategic research efforts in building computational problem solving environments encompass all areas of field physics. The Center operates with faculty, graduate students, and undergraduate students (some 60 faculty and staff researchers and 125 students) from the various academic disciplines, along with a core of full-time researchers, all in a single research facility dedicated entirely to the Center. This has created strong cross-disciplinary interaction among the faculty and students, so that the multi-disciplinary approach naturally encouraged and enabled by high bandwidth connection is now firmly established as a fundamental mode of operation in this Center and as an accepted mode by the academic departments and the University. The success of this Center in addressing engineering applications through computational simulation has resulted in direct involvement with industry (28 government laboratories or agencies and 16 industrial companies) as well as indirect industrial involvement through NASA. And the Center has direct and current involvement with other NSF centers, as well as with DoD and DoE centers. Research and instruction have always been fundamentally linked in the ERC, and as part of its educational mission, the ERC has had over 500 students directly involved in the research of the Center, has developed a cross-disciplinary Computational Engineering graduate program to allow students to integrate their study with the research of the Center, has developed graduate and undergraduate CFS courses for engineering students and others, and has developed a minor in Computational Engineering for undergraduate engineering students. In addition, the ERC has a CFS system for education under development. The ERC has programs with minority and women's institutions, particular Jackson State University and Mississippi University for Women (22 miles away), and works actively with K-12 schools. The Center operates REU programs each summer for undergraduates from other universities and from community colleges, with emphasis on computational science and networked communication. And the Center operates an extensive program of research for undergraduates from MSU throughout the year in these areas. 2. The ERC in the DoD MSRC Program The NSF ERC at Mississippi State took the leadership role in setting up a university team to join with Nichols Research of Huntsville and E-Systems of Dallas to respond to the DoD competitive solicitation for support of the four DoD HPC Major Shared Resource Centers (MSRCs). This university team has the responsibility for the Programming Environment and Training (PET) element of this support, amounting to some $4M @ year for each of the four MSRCs. The ERC led the formulation and development of the PET plan, participated directly in the writing of the proposal, and joined actively in the presentation to DoD. This successful proposal effort represents a strong collaboration between the ERC and industry. This university PET team is as follows: Center for Computational Field Simulation (NSF Engineering Research Center at Mississippi State), National Center for Supercomputing Applications - NCSA (NSF Supercomputer Center at Illinois), Center for Research in Parallel Computing - CRPC (NSF Science & Technology Center headquartered at Rice, including Syracuse and Tennessee), Ohio Supercomputer Center (at Ohio State), Texas Institute for Computational & Applied Mathematics - TICAM (at Texas), HBCUs: Jackson State, Clark-Atlanta, Central State. This university team now has the responsibility for Programming Environment and Training (PET) at three of the four MSRCs: CEWES - Army Engineer Waterways Experiment Station at Vicksburg MS ASC - Air Force Aeronautical Systems Center at Dayton OH ARL - Army Research Laboratory at Aberdeen MD Mississippi State (the ERC) is the lead university, with Jackson State the lead HBCU, at CEWES. Ohio State (OSC) is the lead university at ASC, with Central State the lead HBCU. Illinois (NCSA) is the lead university at ARL, with Clark-Atlanta the lead HBCU. The ERC is also a member of the Northrop-Grumman team - which also includes Virginia, Oregon State, and the San Diego NSF Supercomputer Center - supporting PET at the fourth MSRC: NAVO - Naval Oceanographic Office at Stennis Space Center MS This PET support includes training courses, graduate courses, and side-by-side code migration and enhancement effort in the ten DoD Computational Technology Areas (CTAs), with particular emphasis on scalable parallel environments: CFD: Computational Fluid Dynamics CSM: Computational Structural Mechanics CCM: Computational Chemistry and Materials Science CEA: Computational Electromagnetics and Acoustics CWO: Climate/Weather/Ocean Modeling SIP: Signal/Image Processing FMS: Forces Modeling and Simulation/C4I EQM: Environmental Quality Modeling CEN: Computational Electronics and Nano-Electronics IMT: Integrated Modeling and Testing Support is also provided for Scalable Parallel Programming Tools and Scientific Visualization. These training and graduate courses are provided by a combination of conventional on-site classes, network classes, two-way video classes, and interactive computer tutorials. In addition to leading the university team at the CEWES MSRC, the ERC has the direct responsibility for the CFD support at three of the centers: CEWES, ASC, ARL - and effectively also at the fourth center, NAVO. ------------------------------------------------------------------------------ ----- End Included Message -----