Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Session III What Are the Funding Agencies Doing, and What Are Their Plans for the Future? Session Chairman: Allan Zalkin 77
Computing and Crystallography: The National Science Foundation^and the National Academy of Sciences-National Research Council Peter G. Lykos The National Science Foundation (NSF) makes grants to support innovative projects in the sciences, designed to create or discover new knowledge and, to a lesser extent, to teach and/or disseminate new knowledge. The Research Directorate, one of five comprising NSF, accounts for half the NSF budget. The total NSF budget is surprisingly modest, about the same as that of the Chicago Public School System or about half a billion dol- lars per year. The Chemistry Section in the NSF Research Directorate, accounts for some of the NSF support for chemistry. Other parts of the NSF such as the Offices of Computing Activities and Science Information Service account for a nontrivial portion. Within the NSF, crystallography is not singled out programmatically and explicitly as such. At least three NSF Sections support crystallography, namely, Earth Sciences, Molecular Biology, and Chemistry. In Fiscal Year 71 (July 1, 1970 through June 30, 1971), the grants with crystallography as a principal component were as follows: No. of Grants Total Grants Computer Cost Included Chemistry 42 $1 080 000 $143 000 Earth Sciences 17 480 000 32 000 Molecular Biology 23 780 OOP 98 OOP Totals (prorated for 1 yr.) 82 $2 340 000 $273 000 This does not include crystallographic work done in the Material Sciences Research Laboratories established by the Advanced Research Projects Agency (ARPA) of the Department of Defense, for which responsibility is in the process of being transferred to NSF. There is increasing pressure on NSF to recognize that certain crystallographic techniques have become routine and an established part of science and technology, the conduct of which is no longer eligible for NSF support in and for and by itself. The Office of Computing Activities, part of NSF's Directorate of National and International Programs, has three Sections, namely, Computer Science and Engineering (CSE), Computer Innovations in Education (CIE), and Computer Applications in Research (CAR). I came to the NSF last * The opinions expressed here do not necessarily reflect the policies of the National Science Foundation or the National Research Council. 78
summer to design a fourth Section, Computer Impact on Society (CIS). While doing that, I am functioning as Program Director for Special Research Resources in CAR. In that capacity I shaped two programmatic thrusts, namely, hierarchical computing for laboratory automation, and computer networking to support research. The networking effort is being split off and expanded into a trial National Science (Computer) Network, described at the April 13, 1972 EDUCOM Conference on Networks and Higher Education. Who will administer that network? What computer/communication network technology will be used? When will significant funding be avail- able? Answers to these and other important operational questions need to be known before the deeper problems of computer resource sharing via networking can be addressed. The Office of Computing Activities was the vehicle through which the NSF subsidized the creation and expansion of campus computing cen- ters. It initially had a primary focus on supporting research, but added an increasing, albeit modest,component of educational support dur- ing the life of the Institutional Computing Services (ICS) Program. The ICS Program was terminated abruptly and without warning about two years ago, leaving many university computing centers in serious financial trouble. The following table reveals the annual grant totals for the ICS Program, and for the line items in basic research grants for com- putation: FY '68 '69 '70 '71 No. Basic Research Grants 3917 4146 4041 4679 No. with Computational Support 870 1475 1544 1682 Total Computational Support in Basic Research Grants $4.5M 5.6 5.5 6.5 Total ICS Grants $10.6 M 6.5 6.5 1.6 As grants are made for varying numbers of years, as ICS grants include purchase of equipment which may be operated for as long as five years, and as the computational support in research grants is as listed in the grant and not the audited total, it would be difficult to interpret these numbers on a year by year basis. However, certain qualitative conclusions are clear, namely, that total NSF expenditures for computa- tional support of research (including ICS) has decreased steadily, per- haps by a factor of 2, since 1967, while the number of research grants containing computational support has increased steadily, by a factor of 2, since 1967. The recent survey of computing activities at 2800 colleges and universities by John Hamblen, covering FY 70, reveals that the total expenditures for all computing in higher education was about $472 000 000 79
and that the data collected suggest $512 000 000 for FY 71. Thus of all expenditures for computing in higher education, NSF support accounts for about 2%. Increasingly the true power of the computer, namely as an informa- tion processing machine, is coming to be utilized in support of scientific research. This is reflected in the NSF in the recent reorganization of the Office of Science Information Service (OSIS) which is increasingly concerned with machine-based systems of information storage, handling, and retrieval. Until recently, OSIS concerned itself primarily with scientific literature, and with references thereto. Now the support of projects dealing with storage, handling, and retrieval of scientific data has become a major concern. Indeed those computer applications considered as typical "number crunching" problems include data-base management and file handling as an important, if not as the most important, component. And a large increase in "data pressure", such as availability of the Cen- sus data, and NASA's ERTS-A (scheduled to go up in June 72) and ERTS-B.^ (scheduled for June 73) earth-surveillance remote sensors producing 10 bits of information per day, emphasize the need for considerable research in the hardware/software problems of the collection, reduction, storage, and retrieval of massive amounts of information on a scale hitherto not seriously contemplated. Leaving the role of the National Science Foundation in computing in science, as this Symposium was initiated by the National Research Council's Committee on Computers in Chemistry, it is relevant to examine briefly how the National Academy of Sciences-National Research Council is addressing the subject. About the same time the National Research Council's Division of Chemistry and Chemical Technology established the Committee on Computers in Chemistry, the National Academy of Sciences formed jointly with the National Academy of Engineering a Computer Science and Engineering Board (CSEB). This Board has concerned itself with study- ing and making recommendations on the more global problems posed by the large enhancement in informational technology brought on by invention and proliferation of the information processing machine concomitantly with developments in on-line mass storage devices and ease of interface to our telecommunications systems. Congressman Jack Brooks of Texas has reques- ted $100 000 000 (H.R. 13 200) to enable the National Bureau of Standards to expand its research and development of standards in all areas of in- formational technology. The proposed legislation provides for utilization of the CSEB as an advisory board. On the other hand, the National Research Council in its Committee on Computers in Chemistry has also a discipline-oriented approach to the problem. This Committee was responsible for initiating in 1970 with support from NSF a study of Computational Resources for Theoretical Chem- istry. A report of the first stage of this study was published a year ago. A more comprehensive study of the feasibility and desirability of a national laboratory for computation in chemistry is now in progress with 80
joint support of the Office of Computing Activities and the Chemistry Section of NSF. A similar dual situation is developing within the National Science Foundation. Should the disciplines assume more of the responsibility for computer support in their fields through, say, the support of disci- pline-oriented regional or national facilities, or should a pluralistic approach be adopted where regional or national computer resources are developed, without regard to the particular disciplines or sub-disciplines served? Discipline-oriented centers of national scope are already in existence with NSF support. Among these are the National Center for At- mospheric Research, a facility providing massive computer and other large- scale specialized equipment support to university researchers as well as to its in-house staff, and the Computer Research Center for Economics and Management Science, developing under a five-year grant from NSF to the Na- tional Bureau for Economic Research, which is purchasing its needed com- puter services actually from an off-site supplier. Two computer-communication network developments should be noted. Firstly, public institutions of higher learning in many states are organ- izing themselves on a state-wide basis for computer service and resource sharing. Secondly, commercial computer-communication networks of national scope already exist (Tym-net), or are well along (DATRAN, partial opera- tion scheduled to begin in Fall 1973), or will shortly become available (DoD's ARPA Network already in existence to be transferred within the next two years to a commercial operation). I hope this brief overview of selected NSF and NAS-NRC computer- in-science activities is informative and will aid the crystallographic community in determining what collective action, if any, it ought to take. 81
The National Institutes of Health and Computational Needs and Resources in Crystallography Michael A. Oxman The mission of the National Institutes of Health is to improve the health of the Nation through research, education, and the exchange of knowledge. Because of its mission orientation, there are no formal programs dedicated to further developing technology itself in the basic sciences. Instead, NIH provides financial support for research projects designed to increase our medical knowledge base or to solve specific health-relevant problems. With respect to the field of X-ray crystallography, primary NIH support is through research projects on a problem by problem basis. In Fiscal Year 1972, for example, 207 projects that involve crystallography are being supported (Table 1). Of these, 61 have crystallography as a primary emphasis term with the majority being funded by the National In- stitute of General Medical Sciences and the National Institute of Arthritis and Metabolic Diseases (Table 2). One exception to NIH's emphasis on research projects, which relates to the purpose of todays meeting, is the Biotechnology Resources Program of the Biotechnology Resources Branch (BRB), Division of Research Resources. The BRB establishes and supports regional resources that make available sophisticated technologies on a broad base to the biomedical research community. In addition to providing a service function, each resource is responsible for promoting strong core research and development in its technology, engaging in collaborative efforts between resource core scien- tists and members of the user community unsophisticated in the use of the technology, and providing an arena for training in that technology for both future technologists and biomedical researchers. Areas included in the BRB program are biomedical computing, mass spectrometry, nuclear magnetic resonance spectroscopy, and electron microscopy. The BRB is the NIH focus for such activities, particularly in the area of biomedical computing. Initially, the BRB program supported batch processing systems primarily (Figure 1). However, over the past few years, the trend had been to establish more general-purpose multiaccess systems and process- control systems in order to meet the needs of the maximum number of bio- medical investigators in the most effective way. A few applications have been received over the past few years to establish regional instrumentation centers for X-ray crystallographic stu- dies. For various reasons, none were approved for funding. It appears that there is insufficient demand for the analysis of small, biomedically important molecules that could be performed on a routine service basis. In addition, using presently available technology, large molecules require such an enormous amount of time and effort that the number of analyses that could be performed per year, for example, would be quite small. Since, very few scientists could be served, support from NIH for such a 82
center appears to be unjustified at this time. Finally, the major bottleneck in the area of X-ray crystallography seems to be the avail- ability of appropriate crystals of macromolecules and not the acquisition or processing of data. None of the computer resources funded by the BRB provide any sig- nificant computational support to X-ray crystallographic studies except the centers at Columbia University and Princeton University. Both, how- ever, are rather specialized in that they are oriented toward computer graphics and model building, not large-scale number-crunching operations. Since BRB support bridges many areas of science and technology that re- late to health research, it would be outside the mission of the program to develop a center limited to one specific field of research. On the other hand, as part of their overall service responsibilities, a few centers supported by BRB currently provide limited computational support to crystallographers. Although other centers are not providing such services at present, many have the necessary capacities and capability for doing so. Certainly many general-purpose university systems also can accommodate the crystallographer's computational requirements. In summary, it seems most prudent to take advantage of the tre- mendous computational capabilities that are now available. Many compu- tation centers, especially those based on large computers, are underuti- lized to various degrees. Therefore, selective upgrading of existing systems and developing an appropriate communications network between crystallo- graphy laboratories and appropriate computer facilities could result in a national system adequate to meet the special computing needs of X-ray crystallography. The BRB will continue its efforts to reduce the research overhead of NIH-supported investigators by supporting computers, and is at present gathering a knowledge base to determine what shared biomedical resources will best meet the needs of NIH's clientele. 83
Table 1 Grants Funded by PHS During FY 1972 to Support Research Involving Crystallography, by Dollars NUMBER OF GRANTS PRIMARY EMPHASIS INDEX TERM* Crystallography Computer Analysis 6 Computer Image Processing G Display 1 Computer Programming 4 Computer, Man-Computer Interaction 1 Computer Simulation 2 No Secondary Computer Term 28 X-Ray Structure Analysis 17 X-Ray Diffraction 2 61 TOTAL FY 1972 DOLLARS $ 266,000 153,000 116,000 -0- 71,000 801,000 557,000 148.000 $2,112,000 INDEX TERM*. NOT PRIMARY Crystallography X-Ray Structure Analysis X-Ray Diffraction Neutron Diffraction NUMBER OF GRANTS 35 23 85 3 146 TOTAL FY 1972 DOLLARS $ 684,000 1,113,000 2,035,000 37,000 $3,869,000 TOTAL NUMBER OF GRANTS =207 TOTAL FY 1972 DOLLARS = $5,981,000 *The Division of Research Grants, NIH, maintains a data system on funded programs. Each grant application is coded according to keywords chosen to describe the project, Keywords that describe the basic research effort are listed as PRIMARY EMPHASIS INDEX TERM, Other keywords listed as INDEX TERM. 84
Table 2 Grants Funded by PHS During FY 1972 to Support Research Involving Crystallography, by Funding Unit AWARDING ORGANIZATION National Institutes of Health Allergy and Infectious Diseases Arthritis and Metabolic Diseases Cancer Child Health and Human Development Dental Research Eye General Medical Sciences Heart and Lung Neurological Diseases and Stroke Research Resources Environmental Health Service Air Pollution Control TOTAL NUMBER OF GRANTS FUNDED 11 41 16 1 17 3 83 18 12 203 TOTAL _4 4 85
*â¢â¢ tn 4J c <d M o 14 0) 4J I 5 O 00 o r-4 I o <u 0 M 3, â¢H SNomiw NI suvnoa 86
Computer Support by The Atomic Energy Commission James F. Wagner Let me set the stage for my discussion by presenting a few general sta- tistics on the automated data processing equipment in the AEC. As of June 30, 1971, there was a total of 954 "computers" in AEC's inventory. However, you have to be careful in analyzing this figure. Of the 954, over 750 are smaller computers used for the most part in on-line control and data collection. As a matter of fact, only 201 cost more than $200 000. This level of cost is significant to AEC because most of our management and control effort is spent on those above the $200 000 cost level. AEC's budget for major computers has been about $35 - $40 million per fiscal year. If one looks at the increasing cost for computers, it becomes obvious that if our available dollar were to remain relatively constant the number of computers we could buy would be less. This is one reason for seeking new approaches to acquiring computers, such as the multiple-computer deferred-payment approach which I describe later. As of June 30, 1971, our total investment in automated data proces- sing equipment was about $327 million. Of this amount, $250 million or almost 75% was located at the following 13 major installations: Location Dollar Investment (in millions) Defense Establishments Los Alamos Scientific Laboratory $39 Lawrence Livermore Laboratory 37 Sandia Laboratories (Livermore and 36 Albuquerque) Bettis Atomic Power 16 Knolls Atomic Power Laboratory 15 (We s t inghou s e) Civil Establishments Argonne National Laboratory 21 Brookhaven National Laboratory 15 Oak Ridge National Laboratory 15 Lawrence Berkeley Laboratory 14 Stanford Linear Accelerator 9 Center Bendix Corp. 12 Oak Ridge Computer Technology Center 12 Savannah River Laboratory (DuPont) 8 87
How Much Money Does the AEC Spend in Support In Military-Related Research? Table 1 gives a summary of AEC's operating costs for R&D. A total of $1374.6 million will be spent on R&D in FY 1973. Of this amount, 55.1% is for nondefense-related R&D, representing an increase in the nondefense-related R&D over FY 1971 of about 10% and over FY 1972 of 6.7%. While the defense-related support from FY 1971 to FY 1973 remains relatively constant, the nondefense-related support shows an increase. Total R&D plant construction in FY 1973 will be $1.6 million. Table 1 U.S. Atomic Energy Commission - Summary of Operating R&D Costs (Based on FY 1973 Budget to Congress) (In millions) FY 1971 FY 197237 FY 1973 Percent % of % of % of Change over Amount Total Amount Total Amount Total FY 1972 Total R&D $1302. 9 100.0 $1307 .8 100.0 $1374.6 100. 0 +5. 1 Defense- Related 615. 2 47.2 597 .7 45.7 616.6 44. 9 +3. 2 Non- Defense 687. 7 52.8 710 .1 54.3 758.0 55. 1 +6. 7 Total Research 429.0 100.0 427.2 100.0 454.1 100.0 +6.3 Defense-Related 70.7 16.5 69.5 16.3 77.8 17.1 +ll.9 Non-Defense 358.3 83.5 357.7 83.7 376.3 82.9 +5.2 Total Development 873.9 100.0 880.6 100.0 920.5 100.0 +4.5 Defense-Related Non-Defense 544.4 329.5 62.3 37.7 528.2 352.4 60.0 40.0 538.9 381.6 58.5 41.5 +2.0 +8.3 £/ Includes proposed supplemental totalling $10.0 million. Will the Planned Acquisition of New-Generation Computers at the National Laboratories Provide Excess Computing Power? That can be made available to research workers in universities on a large- scale basis? If so, does the AEC have any specific plans for accom- plishing this or will this be left up to the individual laboratory? The installation of "new"-generation computers at the national 88
laboratories will be the result of what is known as the MCP (multiple computer procurement). Some time back (Fall 1970), one manufacturer proposed that the AEC buy several of the large computers, all to be installed within a six-month period but payment to be spread over four or five years. It was decided that the matter should be pursued further but on a competitive basis. A review of current needs of the various installations was made and it was determined that if the money were available, seven locations could justify the acquisition of a new large- scale computer. Contact was made with the General Services Administration, a request-for-proposal was prepared and issued, and responses from manu- facturers were received. A special task force is currently (April 1972) evaluating the proposals. It is hoped that a selection will be made soon and contract negotiations can begin. The seven locations involved are: 1. Lawrence Livermore Laboratory (LLL) 2. Stanford Linear Accelerator Center (SLAC) 3. Los Alamos Scientific Laboratory (LASL) 4. Argonne National Laboratory (ANL) 5. Union Carbide Corp. - Oak Ridge Computer Technology Center (CTC) 6. E. I. du Pont de Nemours & Co., Inc. - Savannah River Laboratory (SR) 7. Brookhaven National Laboratory (BNL) First of all, if there will be any substantial amount of time on these machines that will be available for use by others, we are not aware of it. The inclusion of these locations in the MCP was based on the need for each of them to increase their computing capability to meet their in-house programmatic requirements. Average monthly utili- zation figures at these locations for their major computers are as fol- lows: Location Computers Avg. Hours Per Month LLL CDC 7600 (2) 696 CDC 6600 (3) 681 SLAG IBM 360/91 560 LASL CDC 7600 585 CDC 7600 603 CDC 6600 630 CDC 6600 624 CDC 6600 518 ANL IBM 360/50/75 668 CTC IBM 360/50/65 553 SR IBM 360/65/ MP 476 BNL CDC 660 (2) 657 89
As can be seen, these machines are used at near capacity. The acquisi- tion of the new computers will provide needed additional capacity. In some cases, the new machine will replace some existing equipment and in others the new machine will supplement existing capability. If it turned out that for a short period of time in the beginning there was some available capacity, we would leave it up to each of the facilities to make the time available to others. There would be no central direction from AEC - at least from the controller's office. How Much Outside Computing is Presently Being Done at the AEC Large Computer Installations? In discussing the "outside computing presently being done at the AEC large computer installations," there are two ways of looking at the question. One is to what extent are we (AEC) using computers located outside the AEC, and the second is how much computing are we providing to other Government users. It must be understood that it is not AEC's policy to provide computer time to non-Government users or for non- Government use. We can provide some dollar figures on the outside time purchase and hourly figures on time provided to other Government users, but not without qualifying them. The number of hours purchased by AEC installations is not practical to provide. Much of this time is for time-sharing and the "hours" are not consistent. Dollars, however, can be provided. These dollars are for time acquired from "commercial" sources. "Commercial" does not include time acquired by universities and other contractors where the computer is used as a part of the performance of some other task. For the six-month period ended June 30, 1971, the following amounts were spent for outside computer services by the various installations: Cost of Computer Time Acquired from Commercial Sources for the Six-Month Period Ended June 30, 1971 Location Amount Sandia, Albuquerque $36 686 Sandia, Livermore ll 639 LASL** 45 983 BNL 937 ANL* 26 294 NAL 33 407 NYU None ORNL 94 890 WADCO 41 925 LBL 1452 LLL* 5428 SLAC 9679 OR-CTC 15 248 90
SR * 1673 Others 354 604 Total $679 935 Installations included in the Multiple Computer Procurement And the following figures show the amount of time (in hours) provided to other Government users for the six-month period ended June 30, 1971: Location Equipment Number of hours Lovelace B5500 ll3 LASL 6600 1205 1401 7600 Sandia, AL 3600 14 BNL 6600 297 INC 360/75 133 NYU 6600 1005 Carbide, CTC 7090 15 360/50-65 Carbide, ORNL 360/75-91 770 LBL (2)6600 964 Total 4516 Economics and Efficiencies of Large Regional Computer Centers The controller's office has encouraged the establishment of "regional" centers or, more accurately, the centralization of computer capability to serve several users. This effort has been acknowledged by members of Congress. Evidence of locations where the type of operation exists in AEC is the IBM 360/75 at the National Reactor Testing Station, Idaho Falls, the Univac ll08 (CSC) and IBM 7090 at the Hanford Laboratory, the CDC 6400 at the Nevada Test Site, the Computer Technology Center at Oak Ridge National Laboratory, the computing center serving Argonne National Laboratory and the National Accelerator Laboratory, etc. All of these "centralized" installations exist as the result of a study made to determine the best way to meet the data-processing needs of a given number of users. Generally speaking, it is probably more economical and more effi- cient to have one or more larger computers than it is to have several smaller ones. You get a greater capacity and it costs you proportionally less. Some people say that for an increase of two in dollars you can get up to eight in increased capability. There are the advantages of 91
larger memory, faster memory, and larger storage. The larger computers have generally better and more efficient operating systems. The controller's office encourages centralized computer facilities for basically two reasons. Overall costs should be less and it is eas- ier to justify the need for capital funds for a computer if several users are to be served. DISCUSSION Lowrey: Dr. Lykos was talking about the establishment of a computer network, and obviously you're going to need people to use it. Will a certain amount of funds, or percentage of funds, be available essentially to try using this kind of computer network? Will one be able to justify computer support for the purpose of simply util- izing and experimenting with the system you're setting up? Lykos: By way of background to your question, the National Science Foundation for something like a dozen years gave money to univer- sities to help build up the university computation centers for re- search and education. That was the computer facilities program (ICS). When abruptly terminated, its annual budget had decreased to somewhat over 6 million dollars a year. That activity, primarily, became the Computer Applications and Research Section (CAR) of the Office of Computing Activities. Because the user community hadn't really assessed its use of the computer for research, and the costs associated with it, there was really no case that could be made for rechanneling those funds in specific ways through the disciplines. Accordingly the CAR budget is substantially smaller than the sum of the budgets of the programs out of which it was shaped. Perhaps I didn't emphasize sufficiently that the big problem is not the adequacy of computer network technology. It's very clear through the ARPA Network, for example, that the technical workability of computer networking is being demonstrated. The main problem facing us is the finding of specific applications that reveal how research can be enhanced through networking. To what extent this can be done 92
within the National Science Foundation's new program is somewhat up in the air at the moment. This new thrust requested impact money similar to what was used last year for the demonstration of computer assisted instruction. As far as I know, the requested im- pact money for this coming fiscal year is not forthcoming. Con- ceivably in the succeeding years laid out for this project, it will be there. Certainly, some CAR support will have to be provided for the purpose of demonstrating the use of computer networks, and that has already been done in the current fiscal year, in three-level hier- archical computing for laboratory automation, for example, and in certain experiments in computer-network-assisted research. But this cannot be done in an extensive way by the Office of Computing Activities alone in the National Science Foundation. Within the NSF, the Office of Computing Activities is a channel between the world of computers and the other disciplines through its Computer Applications in Research Section. Unless some other parts of the National Science Foundation (in particular the Research Direct- orate) become alerted to the computing needs of the scientific com- munity, and unless they reexamine their priorities along this line and act accordingly, then the experiment may very well fail. So, computer-using researchers, the ball is in your court. Williams: Mr. Wagner, I wonder if you could communicate to us the pro- gress being made on the multiple computer procurement with respect to Argonne and NAL? Wagner: I'm not begging off your question, but I'm not privileged with much information. I know just about as little as you do about it al- though I was in on it at the beginning. But right now the AEC is in the process of reviewing the proposals from the manufacturers on the basis of the request for proposals that went out which includes the stipulation on the non-conversion cost consideration. By May 1st a selection will be made, Argonne's included with the other six, and I cannot tell you much more than that. One thing that I should mention. It's essential that we do all we can in that multiple pro- curement to get competition. Although we're accused of being bureau- crats, people don't realize the number of pressures we have. For instance, some vendor will run to Congress claiming that we're not considering his tape units when we buy an initial IBM system. Well, that's a great thing to say but how do you do it? They go to Congress and they go to GAO, and we get complaints right and left. So in the multiple procurement we do our level best to assure competition. As far as the status of the procurement, May 1st is the date for selection and June 15th for the negotiation of contracts. The first delivery, I think, would probably be sometime in the fall, but not at Argonne. The first delivery would probably be at Savannah River. 93
Donnay: If I understood you properly, you told us that a manufacturer who had six machines for sale went to you and made this offer. Is it really the manufacturer who should ask for a computer? Here we have seven outstanding research centers and, from what you say, they didn't ask for the computerâit was the manufacturer who figured out they needed one. Is that the way it should go? Wagner: I didn't cover this point but I said that at all seven labora- tories we reviewed their needs and saw that they needed computers. But we only get 35-40 million dollars a year to buy them. With six or seven laboratories in AEC and on a 35-40 million dollar level a year, we can't buy them all a major computer. What the manufacturer did was say, "Okay, we know you need them but you can't get them be- cause of the budget process. We'll show you how you can." Now, for instance, the program divisions in the Atomic Energy Commission in 1973 asked for 121 million dollars for what they needed in the compu- ter area. We know we can only get between 30 and 40 million dollars. We have never had a computer item cut out of the AEC budget by Con- gress because we're realistic when we go in with the request. But we know we're not getting what we need. All the laboratories had requested them in the budgets but they didn't succeed in surviving the budget cut. When the manufacturer came in and gave us this opportunity, we jumped at it. We already knew we needed them; the laboratories had told us that. The manufacturer just gave us a means of getting them quicker. Freer: Do you anticipate that the 7600 at Lawrence Berkeley Laboratory will be going on the ARPA Network soon? Wagner: I don't know anything about it. There is an internal AEC communications network. To be honest with you, the first time I heard about the ARPA Network was today. Sayre: A few days ago I received a letter stating that the National Re- search Council, Division of Chemistry and Chemical Technology, is sponsoring a study of the feasibility and desirability of a National Laboratory for Computation in Chemistry. The study is divided into 7 sections, of which one concerns chemical structures. Has this study come to the attention of the speakers, and if so can they tell us a little more about it? Lykos: There is a feasibility study under way to look at the question of a national laboratory for theoretical chemistry. That's a sequel to a two-day conference that was held in Bethesda in May 1970, where the general question of computational support for theoretical chemis- try was considered. Even though that was a short conference of two days duration, as a result there emerged a clear sense of direction that this question merited an in-depth study. The feasibility study 94
was put in the form of a proposal from the National Academy of Sciences to the National Science Foundation, was funded, and is in fact under way. The feasibility study has been made more complicated, however, because although it was conceived by chemists, in terms of the prob- lems that chemists are facing today, the Academy's Committee on Science and. Public Policy in reviewing the report of the Bethesda conference charged the follow-up study with all the problems of scientific computing in research in higher education. Those same questions that were posed in Dr. Harvey Brooks's covering letter, by the way, were also sent to the NSF Office of Computing Activities for reaction, and the OCA response is on record. So, the feasibility study is under way with the awareness that some of the questions to be addressed transcend chemistry. There seems to be a general feeling that discipline-oriented centers ought to come into being; accordingly this particular study will be followed with interest by a large audience. The feasibility study group is being led by Professor Kenneth B. Wiberg, a physical-organic chemist at Yale University and a member of the National Academy of Sciences. The study group involves theo- retical chemists, experimentalists, computer scientists, and repres- entatives from academia, government and industry. There will be a first general meeting of the study group on May 5th and 6th to get a first cut at the position papers now in preparation. The final report should be in hand by October. Sayre: The panel on structural studies is due to meet, I understand, on April 24. It would seem appropriate for any conclusions we may arrive at today to be reported to that meeting. Hamilton: I am on that Subpanel, Dave, and my intention is to present a full report of this meeting at that time. Dewar: I must say I have a reaction to the report from the AEC. I find a severe discrepancy between claims to competence in the pro- curement of large computers and ridiculous claims to Congress of 99% utilization. I'd like to ask if AEC, at least internally, has more accurate utilization figures, because this type of estimate makes no sense and is clearly unrealistic. Wagner: The Controller of AEC, Mr. John P. Abbadessa, feels that utili- zation is the key to what we get. What we do about inflated utiliza- tion figures, I do not know. The way the utilization is reported to us is determined by requirements set by the Office of Management and Budget. If you're talking about the figures I quoted from the test- 95
imony to Congress by Livermore, I am not going to argue with you about what they are running on the machine, because I am not in that kind of capacity. If the kinds of things run on the machines should not be run, then somebody's going to have to say that, and I am not about to, because I cannot go to Brookhaven or Oak Ridge or Livermore and look at what they're running and tell them they shouldn't be run- ning it, if that's what you're telling me. Dewar: My concern is that if AEC is in a position where it feels it has to say there is a hundred percent utilization, it may not be in a position to know for itself whether there is time available for pos- sible distribution elsewhere. Wagner: I do not think places like Livermore and Sandia and Los Alamos were pretending that there's a hundred percent utilization, because it's a 24-hour day, 7-day week operation. The only other thing you can say is that some usage is redundant, and that's what I'm telling you I'm not going to argue about. But they are using 24-hours a day, 7 days a week, and those are the ones that were quoted, Los Alamos and Livermore. Larson: We do have at Los Alamos what we'll call a zero priority situa- tion, where if there is free time on the machine a job will be run free of charge. Approximately the first of February I placed a one- hour job on the shelf, and it has not been run yet. Does that an- swer what you're asking about free time? Williams: We have the same thing at Argonne, zero priority. 96
