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.
INTRODUCTORY REMARKS Walter C. Hamilton Crystallographers have long been among the major users of computers in American scientific laboratories. The massive calculations involved in crystal structure determination--least-squares adjustments of hundreds of parameters derived from thousands of observations, and Fourier ser- ies calculated at hundreds of thousands of points â led the crystallo- grapher to early exploitation of the computer. In fact, many of the advances in computer technology were stimulated by crystallographic needs. The purpose of this symposium is to obtain a spectrum of opinion of crystallographers on such questions relating to their computing as whether their needs are presently being met, whether developments in hardware, software, and remote terminals are going to change radically the pattern of crystallographic computing, and whether centralized crys- tallographic computing facilities may play a role. To provide background for the symposium, a questionnaire (see Ap- pendix 1) was circulated to all the approximately 1800 members of the American Crystallographic Association. This questionnaire received some criticism; it is clear that one never knows exactly which questions to ask until the answers are in. There was particular difficulty with Question 3: "What is your real computing cost per year?" Many crystallographers simply do not know, mainly because the amount of the university or departmental subsi- dy is never reported to them. One response reads as follows: "A user on our campus does not know the real cost. Why? On his output he gets a compute cost which is subsidized (or may be if his chairman wishes) up to 87%. The terminal, connect, paper, maintenance, etc. costs (inhouse) are extra and are covered by the department. The user has no know- ledge of the cost of this subsidy." This is of course a problem, but I think that judicious spade work should allow anyone to find out what his institution is spending on computing (including all these extra items) and to determine his prorata share. It behooves anyone doing scientific research to know exactly how much money is being spent in the support of his research. I might note that in an AEC laboratory, the machines have been purchased and the user is charged only operating costsâagain resulting in artificially low values. The total costs, including machine amortization, can however be obtained, and AEC users should be aware of this kind of subsidy.
Now to a few results of the questionnaire: First of all, 142 questionnaires were returned. Since I asked for only one member of each group to return the questionnaire, and since not all ACA members use computers, I would guess that these returns represent about 50% of the laboratories. This estimate is borne out by my own judgment that, among the major laboratories I know, about 50% of the large users were not represented. What Computers Are Crystallographers Using? The results are summarized in Table 1. It is clear that IBM still domi- nates, with the 360 series the major contributor. A few 370 series ma- chines are in operation, and there are still three or four 7000 series machines in use. MDSt of the CDC usage is in the 6000 series, with three or four users of 7600 's and a number still using 3000 series. PDF-10, UNIVAC ll08, and XDS Sigma series also have fair representation. Table 1 Computers Used By Crystallographers Manufacturer Number in use Percentage by respondents to of total number questionnaire reported IBM 92* 52 CDC 44** 25 DEC 15 8 UNIVAC 12 7 RCA 6 3 XDS 5 3 Honeywell 2 1 Burroughs 1 0.5 Electrologica 1 0.5 *Including 68 in the 360 series. **Including 30 in the 6000 series. 10
What Kind of Work Do the Respondents Do? Structure determination 1ll Small Structures 31 Medium Structures 88 Large Structures 21 Other 38 Several respondents defined small, medium, and large structures and were remarkably consistent. Small seems to indicate a molecular weight below 300; large, a molecular weight greater than 800y or 50 to 100 heavy atoms. A few of the respondents are doing very large structures (pro- teins) , although the response from protein crystallographers was much smaller than the 50% quoted above. The other uses of computers by cry- stallographers (Table 2) are extensive but in toto probably represent a small percentage of the total computer time. Many respondents make use of the computer for both structural and other work. Those that do struc- ture determination spend most of their time on refinement (Figure 1). It would seem that a great effort should be spent on making refinements more efficient, a topic discussed later in this symposium. TABLE 2 Computer Usage by Crystallographers Other Than in Single-Crystal Structure Determination Methods research and program development Liquid diffraction: Amorphous radial distribution functions Powder diffraction: Phase and mineral identification Electron microscopy Optical and geometrical crystallography Electron spectra for chemical analysis (ESCA) of solids Thermodynamics of solids Small angle diffraction Membrane structure analysis Diffuse scattering Magnetic studies Anomalous transmission Interferometry On-line control of experiments and displays Supervision and management 11
Type of Computer Use ~^âââ Data Collection, Etc. â â â Structure Solution .. Refinement "Other" 10 20 30 40 50 60 % of Total Computing Time 100 Figure 1 Estimated Time Devoted to Various Types of Computer Use by Crystallographers What Are the Costs of Crystallographic Computing? A summary of the responses is given in Table 3. As noted previously, the figures are difficult to come by. The maximum amount reported was $150 000 per annum. The total of $2.4 M is probably low by a factor of 4. I surmise that the response was about 50% and that hidden subsidies account for another 507o error. Thus, my estimate of the total amount of money spent in the USA on crystallographic computing is about $10 M per year. The $400 000 reported in direct government grants to individual scientists is illuminating. It means, if accurate, that a major share (more than 80%) of the computing dollar is supplied by grants to the central facilities, probably mostly from federal and state governments. 12
TABLE 3 Annual Computing Costs in Crystallographic Laboratories Estimated Cost Range in 1000 $ Number of Respondents Percentage of Respondents 0-5 41 29 5-10 20 14 10-15 13 9 15-20 10 7 20-30 13 9 30-40 11 8 40-50 9 7 >50 8 6 Don't know 16 11 Mean cost: $18 000 Total for responding laboratories: $2.4 M Direct support through research grants: $0.4 M Other direct computing charges: $0.1 M Although crystallographers are becoming more knowledgeable concern- ing their costs for carrying out crystal structure determination, esti- mates (see Table 4) are still hard to come by, as witnessed by the fol- lowing responses: "Estimate is probably +100%." "It depends on the size and method." "We don't have enough statistics to know." Nevertheless, the mean value of $5400 seems reasonable to me and provides confirmation of my estimated total by the following argument. 13
TABLE 4 Estimated Cost per Structure Estimated Cost Number of Respondents $1000-5000 78 5001-l0 000 13 10 001-20 000 7 20 001-30 000 2 30 001-50 000 1 >100 000 1 Mean Estimate: $5400 Figure 2 shows the increase in the number of crystal structure de- terminations over the past few years. There were probably 1500 in 1971. If half of these were in the USA, at $5400 per structure, we derive a to- tal of $4 M. Double this to allow for hidden subsidies, add $2 M for other than structure determinations, and we arrive at the estimate of $10 M stated previously. How Much Will Computer Usage Increase? A summary of the responses is given in Table 5. These are of course sub- ject to considerable variation, but the mean response does not seem sur- prising. There is an approximate doubling every five years as we are able to do more things and have access to more sophisticated hardware and soft- ware. This increase in computer usage does not necessarily imply a 100% increase in computing dollars, for the cost per unit calculation has been steadily dropping. What Experiences Have People Had With Sharing of Information and Facilities? 62% have used remote terminals 37% have used computers other than in their home installations 42% have used program lists 14
There does seem to be substantial amount of such experience, and most active crystallographers are attuned to the feasibility of remote compu- ting. Program lists are not so widely used, possibly because they rapid- ly become out of date. How is Your Work Affected by Limitations of Available Computational Facilities? Unfortunately, I did not include a category, "Not at all", in the ques- tionnaire, but 9% of the respondents wrote this in. If we combine "not at all" with "slightly", we find that 55% are quite happy, 22% are very unhappy, and 21% only moderately unhappy. The nature of dissatisfaction in a few cases is illustrated by the following two comments: 1400 - 1200 - 1000 - 2 m 800 - J. AM. CHEM. SOC J. CHEM. SOC. INORG. CHEM. 600 - 400 - 200 - 1949 1969 Figure 2 Growth in Annual Number of Papers Published Reporting Crystal Structure Determinations 15
Under Explanation of Other Computing: "Debugging newly obtained programs and adaptation to our computers; coping with frequent changes in computer hardware; poor service relationships with the university computer center, especially sheer red tape." Under Limitations on Computing: "But these limitations are pri- marily imposed by security, budgetary and supervisory restraints or constraints; secondarily by the Byzantine complexity of non-scientific non-mathematical, clerical problems of getting on and off the blankety- blank computer." This background information sets' the stage for the prepared papers and discussions that follow. TABLE 5 Estimated Increase in Computer Use Numbers of responses in various categories by 1977 by 1982 0% 23 22 0-257o 28 9 25-507. 22 19 50-l00% 34 15 100-200% 0 6 200-300% 6 0 300-400% 1 3 400-5007. 1 4 500-l000% 2 3 DECREASE 2 2 DON'T KNOW 15 34 16
