Chapter 9

Computing and Applied Mathematics Laboratory



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Chapter 9 Computing and Applied Mathematics Laboratory

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 PANEL MEMBERS Ralph Z. Roskies, University of Pittsburgh, Chair Daniel L. Solomon, North Carolina State University, Vice Chair Marc A. Berger, Georgia Institute of Technology Mary Ellen Bock, Purdue University Allen L. Brown, Jr., Xerox Corporation Pamela G. Doctor, Battelle Pacific Northwest Laboratories Avner Friedman, University of Minnesota Dieter Fuss, Lawrence Livermore National Laboratory Carl M. Harris, George Mason University Samuel P. Marin, General Motors Research and Development Center Andrew M. Odlyzko, AT&T Bell Laboratories Robert E. O'Malley, Jr., University of Washington Jerome Sacks, Director, National Institute of Statistical Sciences David F. Shanno, Rutgers University Norman K. Sondheimer, General Electric Corporate Research and Development Ivar Stakgold, University of Delaware Robert G. Voigt, National Science Foundation Submitted for the panel by its Chair, Ralph Z. Roskies, this assessment of the fiscal year 1994 technical activities of the Computing and Applied Mathematics Laboratory (CAML) is based on a site visit and review of the laboratory's programs by the panel on May 2-4, 1994, and on the 1994 annual reports of the laboratory and its constituent divisions. NOTE: The panel notes that it assessed CAML strategy and programs for fiscal year 1994 in the context of CAML's current mission. The panel was advised at the time of the program reviews that CAML was to be merged with the Computer Systems Laboratory to form the Information Technology Laboratory but was not briefed on the details of the merger.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 LABORATORY OVERVIEW Mission The mission of the Computing and Applied Mathematics Laboratory (CAML) is to develop, make available, promote, and help in the use of the best methods of modern information technology, applied mathematics, and statistics for the NIST staff and CAML's collaborators in industry. Strategy CAML's vision is to provide leadership, resources, and services that support NIST programs and NIST's partners in industry, government, and research institutions. CAML's goals are to achieve barrier-free communications for technology diffusion, credibility of measurement results, efficiency in experimental designs, accuracy and efficiency of models and computation, and communications-based administrative services. These goals support and in some cases underpin NIST's services and cooperation with industry that focus on infrastructural technologies such as measurement and test methods, standards and evaluated data, and the development of generic technologies. The CAML function is primarily to support and collaborate with other parts of NIST; therefore, CAML programs generally serve industry indirectly. Proposed NIST expansions assign specific activities to CAML, but it should be noted that, in each case, CAML works in cooperation with another NIST laboratory that bears the primary responsibility. CAML provides services directly to industry and other government agencies in the sense that generic methods, tools, and software provided to meet NIST's needs are immediately applicable to the needs of industry and of researchers in universities and national laboratories. CAML disseminates information about these methods and responds to requests from external researchers and practitioners to the extent that it can be done without impeding the provision of support to NIST. CAML's staff influences the quality and relevance of NIST's programs through: Research and collaboration. Through extended collaborative research with NIST scientists and engineers from other units, CAML staff members provide needed expertise in the mathematical sciences and computing techniques and thus contribute to their fundamental research, the development of new technologies, and the implementation of NIST services to industry. Consulting. CAML staff is available on call for short-term advisory consultation and assistance in the identification and use of methods from mathematics, computer science, and statistics; the staff also provides a hotline and an advisor desk for assistance with software development or computer performance problems. Through such direct involvement, it identifies needs common to many application areas and directs its research and services accordingly. Technical services. The main elements are the management of the central NIST computing facilities, networking systems, integrated voice-data-video communications systems, and support for administrative applications of information technology. Separate computing facilities are provided for scientific and administrative activities to ensure the security of corporate

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 data and to accommodate the different kinds of services they entail. Advisory and maintenance services are provided along with training in the use of the various devices and software. CAML's strategy is to: Focus on initiatives. The pace of change in computing and communications technologies continues unabated, and the enthusiasm of the Clinton administration for the National Information Infrastructure (NII) has induced interagency committee and planning projects. CAML's director is NIST's Senior Information Resources Management official, and CAML has the opportunity to help formulate and participate in NIST initiatives. NIST's role in the High Performance Computing and Communications (HPCC) initiative is gathering momentum with $1.4 million in new funds for CAML in fiscal year 1994. Also, CAML has been allotted funds for start-up collaborations in NIST 's Advanced Manufacturing and Advanced Materials Programs planned for initiation in fiscal year 1995. This focus on initiatives must be balanced against expectations of other NIST programs for continuing consultations and collaborations. Promote tools and techniques. CAML plans to leverage its resources by providing seminars, workshops, and training for NIST staff, some of which are presented by CAML staff, some by visitors, and some by arrangements with vendors. For computer users, as much help as possible will be made available on-line in the form of “help” utilities, script files, and so on. Continually improve NIST computing and communications infrastructure. For example, CAML has requested funds for upgrading NIST's central computing facilities and expanding CAML's support for NIST's distributed systems. Resources CAML's staff in fiscal year 1994 consists of 227 employees plus 42 guest workers and contractor representatives. Funding totals $33 million, of which $21 million is directly from congressional appropriations for NIST's Scientific and Technical Research and Services (STRS), $8 million from NIST overhead, $2 million from other (federal) agencies (OA), and $2 million from reimbursements for services provided to other NIST operating units. Assessment of CAML Strategy Initiatives CAML's focus on initiatives is an appropriate element of its strategy. CAML's contributions to the HPCC, manufacturing, and materials initiatives were praised by the heads of other NIST laboratories, which rely on CAML for mathematical, computational, and statistical expertise.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Tools and Techniques CAML's products include the development of new mathematical and statistical methods and techniques, as well as considerable expertise in a broad range of applications. The panel emphasizes that CAML's development of these techniques needs to be followed by an extensive dissemination effort beyond the current workshops to assure that they are adopted both within and outside NIST. NIST Computing and Communications Infrastructure The panel supports CAML's continuing emphasis on keeping NIST's computing and communication infrastructure at the state of the art, which also involves a substantial span of mathematics-based technology. Measuring Impact The metrics2 suggested by the Statistical Engineering Division in the CAML publication Description of the Laboratory and Its Operation (CAML, NIST, Gaithersburg, Md., May 1994; p. 44) are excellent ones and should be adopted by CAML in preparing next year's laboratory description. The CAML-wide report to the panel does not mirror the total accomplishment of CAML. Comparing progress as reported by individual divisions with the CAML-level overview suggests a need for a better summary of CAML 's achievements. For example, the work of NIST fellows does not appear in the report at all. Information Technology Laboratory NIST plans to establish a new laboratory, the Information Technology Laboratory (ITL), by merging CAML and the Computer Systems Laboratory (CSL). The panel sees an opportunity for significant synergy in security, networking, and parallel computing and recommends strongly that division chiefs be intimately involved in developing the mission statement and organizational structure of the merged laboratory. The panel sees a danger that mathematical and statistical efforts will be deemphasized in ITL because these are not traditional components of information technology and because it appears to the panel that CSL is less research-oriented than is CAML. The panel believes that 2   Methodology deployment; workshop attendance; consulting load; references; acknowledgments in publications, standard reference materials certificates, and calibration reports; joint publications; invited talks; reprint requests; industry requests; software usage; quality of American Statistical Association/National Science Foundation fellows; participation in research consortia, participation in Cooperative Research and Development Agreements; and contributions to International Organization for Standardization, American Society for Testing and Materials, and American National Standards Institute standards.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 mathematical and statistical activities are essential to the NIST mission and that centralized activity promotes fruitful interactions and professional development and provides a critical mass for outreach and education. The panel also heard strong endorsements from other major laboratory directors for centralized Applied and Computational Mathematics Division and Statistical Engineering Division efforts. The panel urges that the programs of these divisions be continued under the merger and that the new leadership pursue continued focus and project management in applied and computational mathematics. Assessment of Technical Programs The Computing and Applied Mathematics Laboratory conducts research and provides technical collaboration and support to NIST activities and to other federal agency programs in selected fields of the mathematical and computer sciences, including mathematical analysis, statistics, and tools such as mathematical and statistical models, numerical methods, mathematical software and computational techniques, and an on-line catalog system for distributed access to software. The laboratory is responsible for the development and operation of the NIST central computing facilities, which serve all the NIST laboratories; plans and operates communications networks; and supports the design of administrative computing and communications systems. Although most of CAML's technical achievements are addressed in the divisional assessments, the panel highlights the following: The computing divisions (the Scientific Computing Environments Division, Computer Services Division, Computer Systems and Communications Division, and Information Systems Division) appear to be serving NIST well and to have a good sense of future directions. The Statistical Engineering Division's high-quality work has a strong impact on NIST's mission. The Applied Mathematics Group's significant contributions to modeling in materials science are the basis for a Materials Science Center to be formed jointly with the Materials Science Engineering Laboratory (MSEL). The new emphasis on information technology, which will allow NIST to play a much larger role in NII, will require additional expertise in discrete mathematics, theoretical computer science, and parallel computing. Cryptography, multilayer computer communications protocols, and high-level information interchange standards rely on these mathematical disciplines. NIST cannot achieve this expertise without new staff in these areas. The panel also urges CAML to add new vigor by expanding its visitor and postdoctoral fellowship program. Adequate space must be provided for these new people. The NIST-wide Industrial Fellows Program is a very good idea, but NIST will have to develop a reward system that attracts the best candidates.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Recommendations The following are the panel's recommendations for CAML as a whole. In regard to NIST plans to establish the new Information Technology Laboratory by merging CAML and CSL, the panel recommends strongly that division chiefs be intimately involved in developing the mission statement and organizational structure of the merged laboratory. The panel urges that mathematical and statistical programs be continued under ITL and that the new leadership pursue continued focus and project management in applied and computational mathematics. CAML's high-performance computing and communications activities serve the NIST-wide mission. The choice of the appropriate hardware should emerge from consultation between CAML and the other major organizational units, as embodied in the Scientific Computing Planning Team, which has been formed under CAML leadership. The panel urges CAML to add new vigor to its expertise by hiring new staff and by expanding its visitor and postdoctoral fellowship program. CAML Responses to Fiscal Year 1993 Recommendations Given below are some of the panel's fiscal year 1993 recommendations for the Computing and Applied Mathematics Laboratory as a whole (quoted from the fiscal year 1993 assessment), with CAML's responses. CAML should formulate initiatives for assuming national leadership in the development and dissemination of generic technology. . . . ” (p. 246). All areas of the laboratory have been reasonably responsive to this recommendation, and the panel hopes that they continue to be so. For example, the Applied and Computational Mathematics Division did significant work with MSEL on metal solidification modeling, and the computing divisions are working with the Chemical Science and Technology Laboratory (CSTL) and the Physics Laboratory to develop unique parallel-processing applications software. These efforts support other NIST units in their generic technology research. CAML should strengthen its direct linkage with industry by, e.g., assigning CAML professionals to industrial sites, participating in technical reinvestment projects, entering Cooperative Research and Development Agreements, and detailing professional staff to the Advanced Technology Program” (p. 246). CAML strengthened its linkage with industry in both direct and indirect ways. Staff scientists reported a number of direct collaborations as a result of the materials work. Indirect outreach included technology transfer through NIST-sponsored workshops and collaboration with CSTL on chemical process simulation with immediate applications. CAML staff members have served on Advanced Technology Program (ATP) selection boards and as reviewers of ATP proposals, and one is involved in a collaboration among an ATP awardee and two other NIST laboratories.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 CAML should expand the American Statistical Association-NIST-National Science Foundation Fellowship Program to create additional leadership in applied mathematics and computing” (p. 246). The American Statistical Association (ASA) program has been expanded within the Statistical Engineering Division. The Applied and Computational Mathematics Division has opened discussions with the National Science Foundation (NSF) and the Society of Industrial and Applied Mathematics (SIAM). NIST funding for the program has been expanded during fiscal year 1994, and a proposal had been made at the time of the assessment by the ASA to the NSF for continuation of the program. CAML should identify, encourage, and reward team efforts with counterparts in other CAML divisions and other NIST laboratories. . . .” (p. 246). A number of team efforts have been identified, and some have been initiated. CAML divisions should (a) develop strategic plans that appraise the prospects for funding from other agencies, from other parts of NIST, and from allocations for base programs . . . and (b) develop programmatic priorities based on these plans in the course of periodic division meetings” (pp. 246-247). With the exception of the computing divisions, the panel is not satisfied with CAML's development of budgeting options. CAML should join with other NIST laboratories in again exploring the desirability and feasibility of a NIST initiative addressing mathematics-based manufacturing generic technology” (p. 247). CAML is beginning to explore the feasibility of an initiative involving mathematics-based manufacturing with the Manufacturing Engineering Laboratory (MEL). Such a collaboration would seem to be a high-leverage opportunity for all of CAML's divisions. DIVISIONAL ASSESSMENTS Statistical Engineering Division Mission The Statistical Engineering Division seeks to catalyze experimentation, enhance research, and improve communication of results by collaborating with, and developing effective statistical methods for, NIST scientists and NIST's collaborators from industry. Strategy The Statistical Engineering Division reaches out to NIST scientists and engineers through regular seminars, tutorials on important statistical concepts and tools, active involvement in the development of new NIST standard reference materials (SRMs) and calibration services, collaborative research keyed to requests from other laboratories or to ASA-NIST-NSF fellowship proposals, and NIST publications media. Over one-half of the division's activity is in joint efforts with other NIST laboratories. Although the division's primary customer group consists of NIST scientists and engineers, direct interaction with industry has been growing. During fiscal year 1993, division members organized or helped organize six workshops for industrial audiences. In addition, at least six

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 active projects have involved direct collaboration with industry. These projects are conducted in partnership with other NIST laboratories, and many are based on formal CRADAs or research consortia. Some specific objectives of the Statistical Engineering Division are to establish and maintain world-class expertise on the application of statistical methods to measurement science and technology; to identify cutting-edge statistical modeling and analysis techniques and to demonstrate their use in measurement science and related industrial applications; to maintain and continually improve collaborative relationships with NIST measurement scientists; to develop approaches to industrial experimentation that are efficient and robust; and to disseminate newly developed technology developed through workshops and presentations at industrial and scientific conferences. The division's strategy is based on the conviction that the greatest contribution to the NIST mission is accomplished through a competent professional staff.that has freedom to act and is dedicated to customer service within the NIST mission. Realizing such a professional staff culture involves (1) recruiting technically competent and imaginative persons who are committed to the application of statistics through professional collaborations with physical scientists and engineers, (2) nurturing of new staff members by experienced leaders to develop focus on selected customer segments, and (3) experienced leaders' assuming responsibility for a customer group and dedicating themselves to bringing the power of statistics to bear on customer needs. To achieve its technical objectives, the Statistical Engineering Division supports NIST measurement services to industry through the certification of SRMs, the assessment of test methods, and the development of calibration services; participates in the planning and analysis of physical experiments; maintains awareness of industry needs by direct contact with industrial customers and collaborators and by presenting workshops based on NIST experience; and seeks out and develops new opportunities where statistical methodology will have maximum input. Resources In fiscal year 1994, the Statistical Engineering Division, Gaithersburg, Maryland, group, has a staff totaling approximately 20 full-time equivalents (FTEs), supplemented by five guest researchers. In the Boulder, Colorado, group, staff for 1994 consists of about five FTEs. The fiscal year 1994 division budget of approximately $3.2 million consists of about 85 percent STRS base funding, with the balance coming from nonbase STRS funds and NIST interdivision transfers. Assessment of Strategy The Statistical Engineering Division strategy is well thought out and well articulated but should be cast in the context of a set of mutually agreed on thrusts and a management model that ensures focus on those thrusts. The division struggles to balance its role as a consulting service with its desire to take a leadership role in new NIST directions. Individual statisticians within the Statistical Engineering Division appear to accept projects in response to requests with little consideration of division,

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 laboratory, or NIST priorities. Although the criteria offered by the Statistical Engineering Division in CAML's Description of the Laboratory and Its Operation (p. 41) for selection of projects are laudable and appropriate, they do not include a measure of centrality to the division's agreed on thrusts. The division's present organizational structure, in which the statistical scientists are in one unit, offers more efficiency and synergy than one in which they are dispersed among various NIST laboratories. On the other hand, deeper collaborations between statisticians and subject matter scientists can develop when (even informal) liaison relationships are established that link individual statisticians to specific application areas. The balance between consulting and collaboration with other NIST units and direct collaboration with industry through specific projects and workshops is appropriate. Division and laboratory management should ensure that upper administration is regularly apprised of the division's outreach. The morale and dedication of the division may be vulnerable in the absence of division leadership during the impending transition to a new laboratory structure. Putting a new division chief in place may take many months. Interim leadership would be appropriate. The division flourished under its recent leadership. The panel is concerned that statistics (and applied mathematics and economics) is not as natural a fit in the planned Information Technology Laboratory as are CAML's computing divisions. It is therefore particularly important that the division continue to have strong leadership. To attract a permanent division chief of the stature and caliber required to ensure the continued strength of the division, the Senior Executive Service (SES) status of that position should be preserved. Assessment of Technical Programs The Statistical Engineering Division plays a key role both in NIST 's classical mission of precision measurement and in NIST's enhanced mission to improve the competitiveness of U.S. industry. The division ably carries out an enormous workload, involving calibration of SRM and measurement apparatus; statistical modeling and analyses for industry, NIST laboratories, and other agencies; conducting of workshops and conferences, and active participation on standards committees; running a successful and popular ASA-NIST-NSF Fellowship Program; and professional publication and review work. The Statistical Engineering Division's function is multidisciplinary. It not only provides the statistical expertise necessary but also acquires a deep understanding of the physics and chemistry of underlying technical processes. The Statistical Engineering Division supported over 40 SRM projects during 1993, involving calculation of values and uncertainties for physical quantities of interest. The division's primary role is in the design of experiment and testing procedure (e.g., factorial design, complete and incomplete block design, full and fractional design, Youden design), along with the accompanying estimation (e.g., parametric and nonparametric regression, bootstrapping), where statistical modeling is often necessary to capture the effects of systematic and random error. The Statistical Engineering Division brings a wealth of statistical expertise to bear on its measurement work, as well as nearly 20 years of experience in statistical graphics (e.g., its DATAPLOT program). Examples of important programs include measuring latent oxygen concentration in silicon using infrared spectroscopy, measuring optical fiber dimension by digital image analysis of

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 video microscope measurements, measuring radiation exposure by thermoluminescent dosimetry, measuring lead recovery from powdered paint by leaching, and depth profiling to determine boron concentration in multilayer carbon film. The Statistical Engineering Division is actively involved in several long-term metrology projects with industry and other agencies, often through collaboration with other NIST laboratories. The panel notes the particular relevance of these activities to the NIST mission. These projects involve cutting-edge technology. In one study on the accuracy of measurements from coordinate measuring machines, a competence working group was formed involving NIST experts, with the aim of partnership with companies like Boeing, Caterpillar, and General Motors. In another study on measurement of thermal expansion for polymer films used as insulation between different semiconductor layers of integrated circuits, NIST set up a forum of researchers, including participants from the Semiconductor Research Corporation, Microelectronics and Computer Technology Corporation, International Business Machines, and Digital Equipment Corporation. For other agencies, the Statistical Engineering Division is working on the certification of the Charpy machine toughness test initiated by the U.S. Army, and performance analysis of Multiple Instruction, Multiple Data (MIMD) computer programs, under the support of the Advanced Research Projects Agency. The Statistical Engineering Division regularly conducts successful industrial workshops in topics such as mass measurement, analytical chemistry, standards activities, accelerated testing, and statistical methods. These are vehicles for contact and technology transfer with industry and play a significant role in aiding U.S. industry in quality control, manufacturing, and measuring processes, positively impacting NIST's efforts in industrial outreach. The Statistical Engineering Division is in its fourth year of a fellowship program sponsored by the ASA, NSF, and NIST. This program has brought senior and junior academic and industrial researchers in statistics and engineering to NIST and has been an impetus for two-way communication between NIST and the academic research community and the catalyst for recruitment of new young talent. The panel is pleased with this program's results. The Statistical Engineering Division also does extensive review work for other NIST activities, including the ATP, and professional statistical societies in general. Last year division staff published approximately 70 papers in addition to numerous technical reports; gave talks at conferences; and participated in journal editorship, review of papers and proposals, workshops, fellowship programs, professional societies, and other related activities. Two staff members received ASA awards. The Statistical Engineering Division's recruiting efforts, mentoring of new staff members, proactive approach to industrial needs, broad-based familiarity with industrial statistics, and comradeship within the division were commendable. Recommendations The following are the panel's recommendations for the Statistical Engineering Division. CAML should charge an interim division chief for the Statistical Engineering Division to aggressively shape the division's future in the new ITL.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 The Statistical Engineering Division should monitor its selection of projects to ensure that it focuses its effort on those most central to the NIST mission. In developing a project management system, the Statistical Engineering Division should consider establishing liaison relationships organized around high-priority themes. To attract a permanent division chief of the stature and caliber required to ensure the continued strength of the division, the SES status of that position should be preserved. Division and laboratory management should ensure that upper administration is regularly apprised of the division's outreach. Statistical Engineering Division Responses to Fiscal Year 1993 Recommendations Given below is the panel's fiscal year 1993 recommendation for the Statistical Engineering Division (quoted from the fiscal year 1993 assessment), with the division's response. “The Statistical Engineering Division's two-fold mission, involving basic research in statistical science and collaboration with other NIST scientists, should be broadened to include extensive and direct collaboration with industry” (p. 248). The division responded that it has a long list of active projects that involve direct collaboration with industry and that its workshops provide an important vehicle for transfer of statistical methodology to industry. Applied and Computational Mathematics Division Mission The Applied and Computational Mathematics Division provides leadership within NIST in modern analytical and computational methods for solving scientific problems of interest to U.S. industry. Strategy The Applied and Computational Mathematics Division creates a unified analytic and computational setting in which advanced scientific and engineering problems can be solved. The setting includes the development and analysis of theoretical descriptions of phenomena (mathematical modeling), the design and analysis of the requisite computational methods and experiments, the transformation of these methods into efficient numerical algorithms for high-performance computers, the implementation of these methods in high-quality mathematical software, and the distribution of this software to potential clients. The Applied and Computational Mathematics Division provides services to U.S. industries, primarily through collaboration with other NIST laboratories that work with industrial

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 contacts. The division also designs, develops, maintains, and distributes mathematical software that benefits industry. Examples are the software jointly developed with CSTL that enables companies to achieve superior cryocooler system designs; possible licensing agreements for the use of an algorithm for image deblurring; and, in just the first 3 months of 1994, providing access to more than 350 commercial organizations to the division's on-line Guide to Available Mathematical Software (GAMS). The Applied and Computational Mathematics Division also seeks direct industrial contacts as appropriate to strengthen its collaboration with NIST's MEL, increase involvement with the Physics Laboratory and CSTL in the HPCC project, and collaborate in a major initiative with the MSEL. The division assumes that there will be an increasing demand for its services and expertise. Scientists and engineers will have increasingly powerful personal workstations with advanced graphics, parallel architectures, and extensive connectivity. Mathematical modeling and simulation will continue to reduce the number of physical experiments in many areas of science and engineering. Existing computational and mathematical techniques will need further advances in many applications. Staff size is not slated to increase beyond current allocated levels. Permanent staff turnover is not anticipated in the foreseeable future; however, two offers have been accepted for postdoctoral positions, and two faculty persons may spend at least part of their sabbatical years with the Applied and Computational Mathematics Division during fiscal year 1994. Finally, the division is developing a program with SIAM and NSF (similar to the program in the Statistical Engineering Division) to sponsor faculty members for up to 1 year to work on problems of mutual interest. Computing equipment will be continually updated to maintain near state-of-the-art workstations for each staff member, and the division 's budget will increase moderately over the next several years. The division will be moved to an off-campus site to make room for extensive laboratory reconstruction and renovation, causing major disruptions in its activities and creating special problems in remaining an integral part of other scientific activities of NIST. Reinforcing the Applied and Computational Mathematics Division's contact with other laboratory researchers will require a concerted effort by its staff members. Increased frequency of staff seminars and lecture series with outside speakers and short courses in numerical and analytical modeling techniques will help in this regard. Since establishing successful technical collaboration is frequently a matter not only of scientific but also interpersonal compatibility, a major effort by staff members will be necessary to assure that present and future collaborations will remain strong. The division's goal is to be the leader at NIST in the exploitation of modern computing and applied mathematics that will lead to improvements in understanding in the sciences and engineering and in products and processes in industry. Although the division is committed to maintaining broad competence in computing and applied mathematics, it realizes that it must carefully select areas of concentration to be most effective. The division has selected six major technical thrusts. A “champion” has written a plan for each thrust based on input from others who have an interest in the thrust. The plan has been reviewed by each Applied and Computational Mathematics Division staff member. The thrusts are mathematical software, materials modeling, computational geometry, computational physics, parallel computing, and computational metrology.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Resources The division has 24 full-time permanent employees, 5 of whom are located in NIST's facilities in Boulder. Eight faculty members are employed part time, and there are several summer students. An additional staff member was added since the fiscal year 1993 assessment. The division expected to add two additional staff members during fiscal year 1994. Funding includes $2.745 million of base (STRS) funds and $0.450 million from OA and nonfederal reimbursements. Assessment of Strategy The Applied and Computational Mathematics Division's strategy has laudable aims. If this strategy is to be implemented effectively, however, major changes in the operations of the division will be required. The goals are so ambitious—especially with respect to mathematical software, computational physics, and parallel computing —that to achieve them, the division will need to shift toward a much more focused method of operations with several large groups working on well-defined projects and shift away from the present culture in which most researchers pursue their individual interests. For example, the mathematical software thrust has worthwhile goals, but it requires much more effort than is likely to be available. Just to develop expertise in software for scalable parallel computing to the point of being competitive with national supercomputer centers calls for a group of at least half a dozen researchers. There is little such expertise in the Applied and Computational Mathematics Division now, and one hire in this area will not suffice to reach the goal. Similarly, a center of expertise in testing and evaluation of mathematical software could be of major help to U.S. industry but again requires about half a dozen experts to be viable and have a noticeable impact. The Applied and Computational Mathematics Division does not seem to plan to shift enough of its present personnel into this area to achieve its goal. The thrusts in materials modeling and computational geometry are probably the only areas that can be pursued adequately with the resources that are currently directed to them. Assessment of Technical Programs The long-standing intensive and fruitful interaction between the Applied and Computational Mathematics Division and MSEL is accelerating. Among current major joint projects are a study of recently developed phase-field models for problems of phase change, an area in which NIST scientists have a history of major contributions; an industrially oriented study of soldering with both theoretical and experimental aspects; an international effort focused on obtaining a better understanding of directional solidification with electrical effects; and projects covering aspects of polymers and ferromagnetic materials. The division and MSEL are currently in the planning phase for establishing the Distributed Center for Theory and Modeling of Materials. According to its director, MSEL's activities are expanding and the Applied and Computational Mathematics Division's collaboration is welcomed. The impact of the division's ongoing research on materials science modeling is extended through efforts of former postdoctoral associates and others doing related work at universities.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Other valuable research includes large-scale optimization and image deblurring and recognition. The increasing interdisciplinary involvement of the Applied and Computational Mathematics Division staff with scientists throughout NIST benefits from the division's shared experience within CAML and its development of a variety of up-to-date computational and analytical methods and software. The GAMS project has had a dramatic increase in usage as a result of its improved electronic accessibility. It carries on the tradition of the still important NBS handbook by M. Abramowitz and I. Stegun (Handbook of Mathematical Functions, with Formulas, Graphs, and Mathematical Tables, U.S. Government Printing Office, Washington, D.C., 1964). The central importance of numerical analysis must continue to be emphasized, especially in the context of physical modeling. For example, the solution of partial differential equations and boundary value problems underlies the practical success realized in cryocooling and image enhancement. The division's recruitment of a new staff member, an expert in multigrid methods and parallel processing, is to be especially commended. Recommendations The following are the panel's recommendations for the Applied and Computational Mathematics Division. The Applied and Computational Mathematics Division's planning should address the fact that its goals are too ambitious for its current size and structure. The Applied and Computational Mathematics Division should continue to emphasize the formulation of a specific team project in mathematics-based product design and manufacturing to integrate the division's ongoing research in several areas, including computational geometry and metrology, mathematical modeling, and computational science. The Applied and Computational Mathematics Division should be proactive in communicating to NIST management the role and importance of applied and computational mathematics in NIST's interdisciplinary research programs. The division's leaders should maintain a list of accomplishments that have had a significant impact on other NIST operations or on specific industrial practices. Practical impacts should be emphasized. The Applied and Computational Mathematics Division should increase promotion and dissemination of mathematical and computational tools developed by the division. The Applied and Computational Mathematics Division should actively promote and reward participation in the planned NIST Industry Fellow Program and use this as a means to establish direct links to industry. The proposed SIAM-NSF-NIST postdoctoral program should be pursued more aggressively, in addition to increased participation in the National Research Council research associate program and other opportunities for hosting visitors.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 The Applied and Computational Mathematics Division should intensify its collaboration with MSEL through the Distributed Center for Theory and Modeling of Materials currently being planned. Applied and Computational Mathematics Division Response to Fiscal Year 1993 Recommendation Given below is the panel's fiscal year 1993 recommendation for the Applied and Computational Mathematics Division (quoted from the fiscal year 1993 assessment), with the division's response. “The Applied and Computational Mathematics Division should seek more collaborations with other NIST laboratories as well as with extramural partners and focus on a small number of carefully selected options to best serve the needs of NIST and the scientific community at large ” (p. 249). The Applied and Computational Mathematics Division recognizes that this is a major responsibility and that it will require even more attention in light of its impending move off campus. The division made considerable progress in selecting specific projects for collaboration. The strategic plan contains focus areas for the next few years' work. Continued progress is needed to ensure ACMD's success after the coming merger. The continuing promise of applied mathematics needs to be continually asserted and focused. Office of the Associate Director for Computing The Office of the Associate Director for Computing provides NIST with central scientific computing services, network and data communication, software, voice communications, administrative computing services, and related scientific and technical support. Four computing divisions are involved: Scientific Computing Environments Division, Computer Services Division, Computer Systems and Communications Division, and Information Systems Division. Under the Clinton administration, NIST has entered a period of rapid and sometimes dramatic change. These changes include major increases in the Advanced Technology Program and the Manufacturing Extension Program, substantial expansion in NIST laboratory programs, and an emphasis on new methods for accomplishing traditional NIST functions. These changes are having a strong impact on NIST's demand for computing and communication services, and many specific details of needs and plans remain to be completed. Specific activities and resources of the computing divisions follow. Function The Scientific Computing Environments Division provides, trains, maintains, and consults on the scientific software and utilities used on the NIST central computer system; manages and maintains the CAML Ethernet, central server systems, software, and workstations; and develops and supports computer graphics and visualization for central computing and for NIST users' local workstations.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 The Computer Services Division is responsible for managing, operating, and maintaining NIST's central scientific computing resources, primarily in the Central Computing Facility in Gaithersburg The Computer Systems and Communications Division is responsible for NIST-wide computer networks and telecommunications, including planning, design, development, operation, management, and user support. The Information Systems Division supports administrative computing functions at NIST, including those required by statute (e.g., accounting and personnel) as well as those systems needed for effective organizational management (e.g., property, work order tracking). Assessment of Programs Program Planning. The strategy of the four computing divisions in CAML is to focus on initiatives in high-performance computing, promote state-of-the-art computational tools and techniques, and press for the continual improvement of the NIST computing and communications structure. The panel finds that the four computing divisions have made good progress on all of these fronts. The panel commends them on prioritizing support of distributed processing, parallel processing, and visualization and for deemphasizing end-user tutoring, on-site computing at Boulder, in-house applications software development, and supercomputer operating systems support. The computing divisions have positioned themselves to exploit the funds available to the NIST laboratories through the HPCC initiative, coordinating one of the three major components of HPCC at NIST, high-performance software (HPS). HPS is being used to increase the market penetration of mature CAML-supplied tools such as GAMS, help justify the purchase of an SP-2 parallel system to improve the high-performance computational infrastructure, and expand CAML's outreach to organizations both internal and external to NIST by software exchange programs. Also noteworthy is the computing divisions' strategy of outreach to NIST end-users and management by engaging them in planning and upgrade of initiative activities. Finally, the panel acknowledges the computing divisions' initiatives in distributed computing because they help realize strategy. The adoption of standard network operating systems, protocols, and standards should significantly increase their clientele's productivity by enhancing both individual and collaborative productivity. There appears to be a serious problem with the allocation of computing resources; the number of users has grown to the point that completing useful work in a timely fashion on the Cray supercomputer is in question. To deal with this problem, the panel recommends that an allocation committee be created consisting of representative users from across NIST charged with the responsibility of making recommendations for resource allocations based on the technical strength of proposals from the user community. Technical Quality. The Description of the Laboratory and Its Operation (CAML, NIST, Gaithersburg, Md., May 1994) describes the activities of the computing divisions in excellent detail. Overall, all four divisions are addressing appropriate issues in an efficient and productive way. Examples of outstanding programs are tools for managing distributed and parallel

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 computing, centralized software distribution, workstation storage backup, visualization, information management systems, personal computer assistance and maintenance, and efforts to increase the bandwidth in the network, particularly for graphics. The communication tools that provide more user-friendly access to the Internet and World Wide Web or from one's home to the NIST computing facilities will become increasingly important with time, and it is appropriate to be developing this information infrastructure now. The panel is pleased with the computing divisions' interactions with other operating units and through them with industry. The panel also applauds the effort dedicated to the HPCC and in particular to HPS and encourages expanded involvement in the fiscal year 1995 initiatives in the areas of advanced materials and processing, biotechnology, environmental technologies, and advanced manufacturing. Since funding for the next few years is unclear, it is very important to plan for several possible budget scenarios. The panel is pleased that the management of the computing divisions has prepared for options that span the computing power spectrum. The panel also applauds their effort to work with user representatives to define requirements and specify a configuration for upgraded computing systems; all major procurements should be preceded by that process. In summary, the computing divisions appear to be addressing many problems, and activities and plans in place are effective. In general the resources required to accomplish the divisions' projects are adequate, with two exceptions: the parallel programming area appears from the panel's quick appraisal to be understaffed, and the area of computer security deserves more proactive attention. The computing divisions have a primary responsibility to serve NIST laboratories rather than the external world, but resolving the computer security issues would have a tremendous impact on much of the computing nation. The issue of security of networks and computing systems is rapidly becoming a major inhibitor to their broad acceptance in emerging information-intensive applications envisaged by the NII. This affords CAML an opportunity to make a direct contribution to industry by developing and demonstrating network security techniques. The arrival of the SP-2 will offer opportunities for researchers across NIST to explore the impact of parallel computing on their disciplines. The panel applauds CAML's intent to hire support personnel in parallel programming. Recommendations The following are the panel's recommendations for the four computing divisions. The panel recommends that CAML staff investigate the requirements for and plan a network security program. An allocation committee should be created consisting of representative users from across NIST charged with the responsibility of making recommendations for resource allocations based on the technical strength of proposals from the user community. CAML should consider adding a PhD-level researcher in computational science with expertise in parallel computing to fully utilize the SP-2.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 Computing Divisions' Responses to Fiscal Year 1993 Assessment During its 1993 assessment the panel suggested that the computing divisions focus attention on one or more projects at the forefront of contemporary computing practice which would allow the divisions to contribute to the national effort in high-performance computing and communication. Examples of possible projects included (1) networked workstations based on tools like PVM, LINDA, or CONDOR, which would lead to the development of industrially relevant expertise on production computing and the distribution of useful codes; (2) information management systems based on extensions of Gopher and X-Mosaic (which can include the Guide to Available Mathematical Software) to facilitate the electronic dissemination of NIST's databases and address problems of the metadata as related to the sharing of software; and (3) applying mathematics in areas such as materials modeling, chemical design, and process optimization on massively parallel processor platforms. To their credit, the computing divisions undertook initiatives in all three of the suggested areas. Work is currently under way with the Physics Laboratory to evaluate the efficiency of using networked workstations operating in parallel as an alternative to the Cray supercomputer on a variety of computationally intensive problems. Results to date are incomplete, but the project is an interesting and appropriate use of the computing divisions' resources. Mosaic presents an interface to GAMS, which has resulted in greatly increased query activity. To explore the effects of massively parallel processing for appropriate problems from the NIST environment, CAML is acquiring a 19-processor IBM SP-2, with plans to upgrade the computer to 256 processors when funds become available. In addition to efforts on all three of the above suggestions, the computing divisions have developed a multimedia electronic environment for communication, data exchange, and other services, fully in keeping with the recommendations of the 1993 panel. In summary, the response to the panel's suggestion has been laudable. Office of Applied Economics The Office of Applied Economics (OAE) is making important contacts with industry in the area of decision-making science. One important example is its development of the AUTOMAN expert choice software. The panel did not have the requisite expertise to fully assess its program but believes that OAE activities should continue and that management should seek to stabilize OAE resources by rectifying its overreliance on OA funding. Recommendation The following is the panel's recommendation for the Office of Applied Economics. Management should seek to rectify the Office of Applied Economics ' overreliance on OA funding.

OCR for page 227
An Assessment of the NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY PROGRAMS: Fiscal Year 1994 This page in the original is blank.