196704
196704 196704
User Manual: 196704
Open the PDF directly: View PDF .
Page Count: 60
Download | |
Open PDF In Browser | View PDF |
£Tt~6 LOZIN* +JOTO 11*7L*7 £9 OGNV ~J ~3~ "'WT ' J3S ~sor VS + April, 1967 NVS 13~~VW Sl'!fJIOOI~3d • '7? 0 "y CD 1 ~ CD t Mirror {PI" I~ 1/ by l Computer Why not? -because The Kelly-Springfield TI " Com pa ny gets the right ti re to t;; right place at the right time, with:-, modern communications and d2:;1 processing system, that's why nt,; And they sell over 2600 differe sizes and types of tires in the U.' Canada and 66 other countries. Here's how the system works: each warehouse Bell System Te type® machines use master ta~ to enter standard information ,i customers and their purchasls, while the variable sale informatit n is typed in manually. The machines make printed orders and by-product " ~.. p ~ ! n c h~; d P -J ' ,r:: ',_ 'f) f:; " i"inittea c,. ;2: : :."~:,: "~(..i:; a minu~e. i-,'; "L, fJU L110 U I -::1',"::: :::.' ;1lUed in E; Wc~(;kly ;:;; JC :: .; t " n~,;' ': .t,,,;,.. 1:_; '(' ,:::~ ,;;:- ;i':!"l: (C:-:Y)rl~; !. (", ('Hr~;'lt~i.:"l. '~. t ,., ',"'is, sa,·:, ',-, nei ordc rs ' hC\!:~'2 in thc~- '';'',.'. (;'C'imr.-y. '" ',' 'me;!'\,' :'; ... ---::-l I "peed ~. ,>Ietn· r':'~~:~ --.. :;;,1. ,I I, S "0';'" . '\TS Wide A I<-':Jilone SE . ,......} .. 1 ' 't '.,::.- ... r.l..' -.-. ..:..:. .. ......... _-- ,'"r , . ~ .,:. : :.~.' , (:, ,tel from each of the 26 warehouses i>: rnepared for factory production ~' : ,: juling. TI ~ resu It: I nformation flow is f0~: r, more accurate. Customers e':, ; 1e best possible service, every,,'1';':' e in the Kelly-Springfield marh_ i' g world. ; "i' more information, call your l'~':;\l Telephone Business Office and ()~;I( for a talk with one of our Comrnu'.ications Consultants. And don't Il';~;' :ate to call us early-because if. :; way we can serve you best. \' ,len you work with data communi:'""tions, work with the Bell System. Introducing the DATA 620-1 new systems computer The DATA/620-1 integrated circuit computer is the newest member of the DATA/620 family of system computers. DATA/620-1 fills the gap between general purpose and special purpose computers. It belongs in a system, and solves problems previously considered too difficult or expensive for computer solution. Designed for faster problem solution the DATA/620-1 has a bigger instruction set, integrated circuit reliability, is smaller, has one-half the components, and costs less than any computer in its class. DATA/620-1 comes complete with software, field-proven and refined on the DATA/620. Extremely compact, the DATA/620-1 requires only 10" of 19" rack space. It's available with memory modules from 1024 to 32,768 words of 16 or 18 bits, and with a selection of control, arithmetic and I/O facilities, including D.M.I.'s unique Micro-Exec. Price: $13,900 with 4096 words of 16 bit memory, including ASR 33 teletype. We are very proud of our new DATA/620-1, and would like to tell you more in a fact filled brochure. Please write for one. DATA MACHINES 1590 Monrovia Avenue, Newport Beach, California Tel. (714) 646-9371 TWX (910) 596-1358 Division of DECISION Control, Inc. Designate No. lOon Reader Service Card The Data 620-1 and the Data/620 will be on display at booths HS-HIO at the SJCC. Stop' by and discuss your digital system requirements with a Data Machines applications engineer. The 20-minute education break. For busy people who need to know. Maybe you have a programmer who needs to brush up on COBOL, or a tab operator who needs to know about the IBM 85 Collator. Or you may know a junior executive who just wants to learn a little bit about data processing. IBM has an easy way to teach them. It's called programmed instruction - P.1. for short. P.1. is an approach to self-study that guides a student through the required material at the student's own pace, a step at a time. He can breeze through five or ten pages whenever there's some time to spare-at home or on the job. Last year, some 100,000 students took IBM programmed instruction courses. And the results were gratifying. Students learned faster- 27% faster on the average-and scored higher than did those in conventional lecture-discussion classes. Right now there are 29 P.1. courses available to IBM customers covering punched card systems, computer fundamentals and computer programming. More are being developed. But programmed instruction won't replace your local IBM Education Center. Some subjects are just too complex or too specialized to be taught by P. I. So we'll continue to operate our school system-providing a comprehensive curriculum for everyone from key punch operators to board chairmen. Either way-sending the student to school or the school to the student-IBM education helps you and your people learn how to get the most productive work from your IBM system. IBM Education: it's there when you need it. Just like the rest of IBM's services. ©CQ)[Flfl)CQ)(1)T~~~~ S[nHo'l alUJ~Oll'il'ls'tCaIOJfn Mirror by Computer The front cover shows a famous Japanese woodblock print, as reproduced by a plotter guided by a digital computer. For more in/ormation, see page 60. APRIL, 1967 Vol. 16, No.4 Special Feature,' 13 c&a problem corner by Walter Penney, C.D.P. 37 calendar of coming events 39 across the editor's desk Computing and Data Processing Newsletter 54 computer census 56 books and other publications 59 new patents Performance Comparisons of Computing and Data Processing Systems 16 REAL-TIME PROCESSING POWER: A STANDARDIZED EVALUATION, by J. Burt Totaro How to obtain valid measurements of the power of a computer to. do real-time processing. .22 COMPUTER PERFORMANCE PROJECTED THROUGH SIMULATION, by Fred C. Ihrer A specific evaluation technique predicts the performance of various hardware and software designs in terms of cost and response_ 29 RECOVERY FROM ERROR, by Jan B. Hext The emphasis should be on the prevention of error, as hardware becomes more complicated and more reliable. But software failures seem destined to survive longer. 34 LANGUAGE ENGINEERING, from Hughes Aircraft Company To date no one has assembled a complete, integrated, automatic machine that listens and speaks. But as such devices become more and more needed, advances in making and pro_ gramming them should culminate in a satisfactory, working system. by Raymond R. Skolnick 60 ad~ertising index editm'ial Time Sharing, Batch Processing, and Direct Access editor and publisher market report Edmund C. Berkeley The top 100 computer users (among industrial firms) with over $2 billion worth of computers workinq in 1966 - and who they are assistant editors Moses M. Berlin, Charlene J. Hofer, Linda Ladd Lovett, Neil D. MacDonald world report - market research director Patrick J. McGovern capital report, The Defense Department seeks bids for a management information system; the National Data Center controversy continues; in an experiment, computer systems successfully match unemployed workers with jobs. contributing editors John Bennett, Andrew D. Booth, Dick H. Brandon, John W. Carr, III, Ned Chapin, Alston S. Householder, Peter Kugel, Rod E. Packer advisory committee Great Britain, Kosygin praises British computers; sales to East Europe rise; British universities purchase U.S_made machines; British European Airways to purchase integrated management system. multi-access forum 8 Time-Sharing: A View from New Zealand, by B. A. M. Moon Perhaps the benefits of time-sharing may be available without time-sharing. 9 Library Service for Blind Computer Programmers Seeks Type-Composing Tapes 9 Addresses of Computer Organizations 9 Modern Indian Story: Electronic Firms Locate on Reservations Aid from the Department of the Interior and an abundant, easily-trained labor supply leads to industrial growth on Indian reservations. 10 Cybernetics in the U.S.S.R., by Robert W. Brainard and William D. Hitt Computers in Soviet law, medicine, education, public health, and economics may result in fundamental changes in Soviet economic theory and practice. William J. McMillan, 815 Washington St., Newtonville, Mass. 02160, 617 DE 2-5453 11 Computers Checking All United States Taxpayers for First Time, by Norman Teich The Internal Revenue Ser.vice's nationwide computer system. advertising representatives 11 IFIP Admits Hungary as 26th Member Country NEW YORK 10018, Bernard Lane, 37 West 39 St., 212-BRyant 9-7281 CHICAGO 60611, Cole, Mason and Deming, 737 N. Michigan Ave., 312-SU 7-6558 12 "Individual Privacy and Central Computerized Files" - 12 "Bootstrapping a Career in the Computer Field" - 12 ECPI Program Write-Ups Available LOS ANGELES 90005, Wentworth F. Green, 300 S. Kenmore Ave., 213-DUnkirk 7-8135 ELSEWHERE, The Publisher, 815 Washington St., 617-DEcatur 2-5453 Newtonville, Mus, 02160 13 Jovial Bulletin Established - 53 Revived "Annual Review in Automatic Programming" - 53 Correction 58 Articles for "Computers and Automation" and how to write them T. E. Cheatham, Jr., James J. Cryan, Richard W, Hamming, Alston S. Householder, Victor Paschkis art director Ray W, Hass fulfilment manager Comments Comments *0 ON YOUR ADDRESS IMPRINT MEANS THAT YOUR SUBSCRIPTION INCLUDES THE COMPUTER DIRECTORY. *N MEANS THAT YOUR PRESENT SUBSCRIPTION DOES NOT INCLUDE THE COMPUTER DIRECTORY. SEE NOTICE ON PAGE 20. Papers Sought Papers Sought ..~\ COMPUTERS AND AUTOMATION IS PUBLISHED MONTHLY AT 815 WASHINGTON ST., NEWTONVILLE, MASS. 02160, BY BERKELEY ENTERPRISES, INC. PRINTED IN U.S.A. SUBSCRIPTION RATES: UNITED STATES, $15.00 FOR 1 YEAR, $29.00 FOR 2 YEARS, INCLUDING THE JUNE DIRECTORV ISSUE; CANADA, ADD SOc A YEAR FOR POSTAGE; FOREIGN, ADD $3.50 A YEAR FOR POSTAGE. ADDRESS ALL " . , / EDITORIAL AND SUBSCRIPTION MAIL TO BERKelEY ENTERPRISES, INC., 815 WASHINGTON ST., NEWTONVILLE, MASS., 02160. SECOND CLASS POSTAGE PAID AT BOSTON, MASS. : '~~~r. POSTMASTER: PLEASE SEND ALL FORMS 3579 TO BERKELEY ENTERPRISES, INC., 815 WASHINGTON ST., NEWTONVILLE, MASS. 02160. © COPYRIGHT, 1967, BY BERKelEY ENTERPRISES, INC. CHANGE OF ADDRESS: IF YOUR ADDRESS CHANGeS, PLEASE SEND US BOTH YOUR NEW ADDRESS AND YOUR OLD ADDRESS (AS IT APPEARS ON THE MAGAZINE ADDRESS IMPRINT), AND ALLOW THREE WEEKS FOR THE CHANGE TO BE MADE. COMPUTERS AND AUTOMATION, FOR APRil, 1967 5 c&a EDITORIAL Time Sharing l Batch Processing l and Direct Access In this issue in Multi-Access Forum we publish an important and interesting letter from B. A. M. Moon, Director of the Computer Centre, University of Canterbury, Christchurch, New Zealand. Among other questions that he raises are: 1. If immediate man-machine interaction in scientific and technical research and design problems is so· important, why has there not been more of it in the past? 2. If small computers are going to be needed at the terminals of large time-shared machines anyway, would not the greatest economy in many cases be simply to dispose of the central machine and leave each user in intimate contact with his own smaller computer? 3. Can we get the benefits of time sharing without time sharing? These questions have points of resemblance to the following questions: "Are apples or oranges better"? (Answer: It depends on many factors.) "Is it better to have a pain in your head or a pain in your foot?" (Answer: The answer depends on two factors essentially: if the pains are equally intense and equally serious, certainly the foot pain is preferable.) "Is it better to have the right equipment or the wrong equipment to do a job?" (Answer: Of course, the right equipment is better though in this case, the issue has been avoided by inserting answer-begging words in the question.) Most of the answer to Question 1 is that as long as computers were very expensive, it cost too much to give the persons working on problems in design and research direct access to computers. However, as c;osts have gone down - due to both time sharing and the arrival of small computersdirect access (man-machine interaction) became economically feasible and was immediately desirable, because of latent but ineffective demand. The answer to Question 2 is clear. If most of time sharing reaches the point where a small computer is needed in the console, then clearly we should get rid of the big computer, the intricate software for control, and the long communication lines for access, and we should in nearly all cases settle either for single-console access to a single computer, or for two or three nearby consoles, short communication lines, and simple supervising software in the computer. I am convinced that the answer to Question 3 .is: In a great number of places, we can get the benefits of time sharing without time sharing; the pendulum of that fashion has swung too far. COMPUTERS and AUTOMATION for April, 1967 In fact, we could rename much of "time sharing" as "direct access" and have the substance of it. We should translate a great many of the arguments for time sharing into arguments for direct access. Nevertheless we should remember that: • Batch processing is almost certainly the right procedure for a great many jobs, such as production runs of large computations according to established needs and fully reliable programs. • Time sharing of a powerful computer and a large file of information, centrally located, and with long communication lines, is vitally necessary for some problems - probably not many - such as the airlines reservations problem. • Direct access is a desirable, and economically feasible, facility for the programmer who is troubleshooting a program, the engineer who is experimenting with different ways of solving a problem, and some other people. In fact, we can foresee not a cleavage between time sharing, batch processing, and direct access to single computers, but a mingling. The time is coming when normally the big computer will have a load of batch processing jobs for background, and direct access in the foreground, not only for the operator, but for one, two, three, or even a dozen input lines from various nearby consoles and small computers. The resi.dent software supervisor in the big computer will find this hardly any more trouble than to meet the requirements of twenty input lines from magnetic tape units or similar peripheral input/output devices. There is a phrase I like to remember from the book Silas Marner by George Eliot, which was said by the landlord of a tavern, who used it to settle arguments among his contentious patrons: "Ye're both right and ye're both wrong, as I allus say, - the truth lies atwixt and atween you." The right solution to a particular problem of a computing facility consists of considering all the important factors, and making a sensible well-reasoned choice, with little regard to fashion. Editor 7 c&a MULTI-ACCESS FORUM riME-SHARING: A VIEW FROM NEW ZEALAND I. From B.A.M. Moon, Director University of Canterbury Christchurch, 1, New Zealand Somewhat detached here from the principal centres of investigation, we have watched with interest the growing activity in the use of time-shared computer systems, helpfully reviewed in recent issues of Computers and Automation. In using computers in a scientific or technical environment, the main reason advanced in support of the timeshared system seems to be the opportunity for man-machine interaction in the solution of design or research problems which it offers. In some situations there are further advantages, such as the opportunity for dynamic interaction through the computer between research workers with common problems. At present the remote console is usually a typewriter, considered to be a limiting factor in many cases, though perhaps more apparent than real because we have become accustomed to the sorts of high-speed input/output devices needed with batch-processing systems. Nevertheless, in writing of the next step, Hal Becker (Computers and Automation for Oct., 1966) suggests the provision of card readers, line printers, and even small computers at the remote ter'minals. All this raises the cost, an important point noted by Hans Jeans (also Computers and Automation for Oct., 1966), who observes further that so far in time-shared systems "cost is seldom emphasized as a design criterion." While the well-known "Grosch's Law" indicates that there are economies in the use of large-scale central processors, the increased cost of terminal devices and terminal servicing complexity, which may increase exponentially with the number of terminals, indicate that overall the cost advantage may not lie with the large system. Jeans, arguing further, suggests that the modest system may be no more than most subscribers will need (at least perhaps for those kinds of situation where intimate manmachine rapport is important), a case advanced further by both Rudy Stiefel and George Zimmerman (Compute.rs and Automation for Nov., 1966). If this be true and if small computers are going to be needed at the terminals of large time-shared machines anyway, perhaps the greatest economy in many cases would be simply to dispose of the central machine and leave each user in intimate contact with his own smaller computer. The really big computational tasks would continue to be run on really big computers by bat~h processing methods. 8 An interesting question at this point then, is: If immediate man-machine interaction in scientific and technical research and design problems is so important, why has there not been more of it in the past? T.he opportunity has existed for a number of years; in fact the IBM 1620 and some other machines were designed expressly for this purpose. The cost has not been prohibitive (and may decrease), compared with the cost of a share in a large time-shared system. Yet our experience, at least, with a 1620 has been that users do not desire to use it in this way. Even when given every opportunity for "open-shop" use, they appear much to prefer to enter source statements on prepunched cards, rather than at the console typewriter and to design production runs for lengthy calculations with a minimum of operator intervention. For this reason, we intend to install shortly an IBM 360/44, a fast batch-processing system with even less operator intervention. Does this mean that the attractiveness of a time-shared system in which each user has his own console typewriter is a psychological rather than a technical one, or did computer users simply fail to appreciate what the 1620, say, had to offer? When our 360/44 is installed and the 1620 enters a merited semi-retirement we intend to investigate this point - to initiate a project in whi~h we approach its console typewriter as if it were one of the fifty such typewriters attached to a time-shared system fifty times as fast. In this way we may get the benefits of time-sharing without time-sharing, or else discover that time-sharing has something else to offer which is fundamental and important. Do we seem naive or does the distance of our view give it at least a little perspective? II. From the Editor Thank you for your letter, which I was very glad to read. I think you have raised some important questions, which have served as the basis for our editorial in this issue. Personally, I think "time sharing" expresses the wrong emphasis," and that "direct access" expresses the right em. phasis. I have found in a year's personal use of a timeshared console to a big computer much frustration, and in several years' use of "direct access" to a small computer much satisfaction and only a small amount of frustration. COMPUTERS and AUTOMATION for April, 1967 TECHNICAL LIBRARY SERVICE FOR BLIND COMPUTER PROGRAMMERS SEEKS TYPE-COMPOSING TAPES I. From David L. Neblett Proiect Director Computer Training for the Blind Washington University 724 South Euclid Ave. St. Louis, Missouri 63110 II. From the Editor Weare organizing a technical library service for blind computer programmers. This service will eventually provide technical articles from various sources in Braille to enable these individuals to keep abreast of this dynamic field. We are using the IBM 360 computer to make the translation into Grade II Braille. (Grade I is letter for letter and Grade II is similar to shorthand notation.) I would like to know if you use composing tapes or any similar method in printing that could be used for computer input. Would you make them available to us? In any event we are interested in subscribing to "Computers and Automation." Please enter a subscription and bill us. Thank you for your subscription and for your letter telling us about your service for blind computer programmers. We are sending you some of the punched paper tape produced at our typesetter's, Wellesley Press, One Crest Road, Wellesley, Mass., 02192, for typesetting. They and we will gladly arrange that tape like this be sent to you each time our magazine is set, if this kind of tape actually meets your requirements. Please let us know. ADDRESSES OF COMPUTER ORGANIZATIONS I. From Hans Lutke, Engineer 117 Berlin German Democratic Republic Your editorial on "The Thirst for Computer Knowledge" in the outstanding Computers and Automation for July 1966 encourages me to ask the address of the International Federation of Information Processing Societies. I am interested in their glossary related to computers. This glossary would be of considerable value in my work, in either the English or German language. Please be so kind as to' deliver my wish to said Federation or let me know their address. In either case, I ani much obliged to you. II. From the Editor Thank you for your letter. The address you wish is: International Federation for Information Processing, 345 East 47th St., New York, N.V. 10017. We believe they will be able to help you if you write directly to them. III. To our Readers As a regular service to readers of Computers and Autom.ation anywhere in the world, we are glad to provide the address of a computer-related organization, if we know it. Any request should be addressed to the editor. If you are a part of a computer-related organization, and you are not sure that Computers and, Automation knows the address, please send it to us. In this way we can help increase communication and cooperation in the field of computers and data processing. MODERN INDIAN STORY: ELECTRONIC FIRMS LOCATE ON RESERVATIONS (Based on a report in the Feb. 27, 1967 issue of Electronic News, published at 7 East 12th St., New York, N.Y. 10003) It is a safe prediction that more electronics firms will be springing up on Indian reservations. One reason for this is that the Department of the Interior has been nudging these firms a little with one hand, and offering them sweeteners with another. Social factors are also a consideration. The most important fact, however, is that electronics companies which have pioneered in employing Indians have already demonstrated that they can produce and deliver. Other types of firms are als() showing interest. Interior Secretary Stewart L. Udall has briefed industry representatives on the advantages of operating in Indian areas and of cooperating with the Bureau of Indian Affairs IndusCOMPUTERS and AUTOMATION for April, 1967 trial Development program. He has shown that Indians are easily trained for skilled operations, and represent an abundant labor supply. The Bureau provides on-the-job training programs under which the Bureau contracts to pay up to 50 percent of trainee wages in plants on or near reservations. Indian trainees are recruited and screened by specialists ip the Bureau in cooperation with local offices of the State Employment Service. The Bureau also offers services in getting firm and tribe together. Various tribes also put up money to help finance plants. To date they have set aside more than $12 million to be used for loans for this purpose. 9 A typical experimental plant on a reservation is the Fairchild Camera & Instrument Semiconductor Division located on a Navajo reservation at Shiprock, New Mexico. The firm "employs about 450 assemblers and mechanics, and has a $700,000 contract with the Bureau of Indian Affairs to provide , on-the-job training. The firm is negotiating with the Navajo to build a $1 ~ million permanent facility there, which however is contingent on the provision of proper housing by the Government. Dr. Robert Noyce, group vice-president of Fairchild, said the company regards the experiment as successful bcause "in the Indian culture there is an emphasis on producing when you work!" He admitted labor cost was a factor in the decision to employ Indians, but declared social benefit was also a consideration. "Probably nobody would ever admit it, but I feel sure the Indians are the most underprivileged ethnic group in the United States," he said. A major problem is educating large groups of people on the reservation. The people are so widely dispersed, with cities mere crossroads and without centralized housing, that it is difficult for a large industrial enterprise to employ a great number of people in one location. In another experiment, a group of Seminoles are devoting their talents to assembling components at a $500,000 Amphenol Connector division plant near Hollywood, Florida. The plant is expected to ship approximately $7 million worth of computer, instrument, and communications equipment connectors this year, with an expected increase to $10 million in 1968. As an outgrowth of these successes, an Employment Training Center for Indians will open this month in Madera, California. It will be directed by Philco-Ford Corp. under a contract with the Bureau of Indian Affairs. This Center is aimed at alleviating chronic unemployment and hard-core poverty among Indians, and is a major departure from other programs to increase opportunities for the unemployed. The concept involves comprehensive family training. The center will provide education, training, urban adjustment orientation, and job-placement for all family members of employable age. Family and vocational counseling, pre-vocational and ~ocational training, and health services, will be provided for trainees and their families, at no cost to them. Specialized training for jobs, job placement, and follow-up, are included. Philco-Ford is committed to place all trainees who complete the program. It is expected that more than one job will be generated for each family. The center will initially enroll 30 Indian families from all parts of the country. An additional 200 single trainees will be added by July 1. Theodore S. Hoffman, president of Hoffman Info'rmation Systems, Inc., has been appointed a consultant on economic development with the Bureau of Indian Affairs. He will advise the Bureau on development and use of Indian talents and resources; promotion of locations on Indian reservations for industrial plants, commercial endeavors, and tourist enterprises; encouragement of Indian-owned-and-operated commercial ventures; and mobilization of credit and financing for such activities. CYBERNETICS IN THE U.S.S.R. Robert W. Brainard and William D. HiH Battelle Memorial Institute Columbus, Ohio 43201 The Soviet Union is making a considerable effort to integrate cybernetics with the beliefs and institutions of Soviet society. In the Soviet Union cybernetics has kept a solid scientific meaning as the science of control and communication in man, machines, and organizations with emphasis on the use of automatic computers; and the subject is being applied to a very broad range of activities, including factory production control, economics, law, medicine, and education. The Soviets consider that cybernetics is relevant to contemporary scientific and technological problems, and reinforces certain statements of dialectical materialism, a' philosophy of development of societies and similar complex phenomena. Official recognition, and acceptance of cybernetics in the Soviet Union began in 1958 with the establishment of the Scientific Council on Cybernetics by the USSR Academy of Sciences. The most ambitious application is in the field of economics, where a national program aimed at establishing a cybernetic system for controlling the entire economy is being vigorously developed. At the heart of this proposal is a dynamic mathematical model of the entire Societ economy. Planning and control would be implemented through an integrated nation-wide network of computer centers. Soviet experts predict that this program, when introduced, will raise the national economic output by one-third, and will double the present rate of economic growth. Introduction of cybernetic methods could result in fundamental changes in Soviet economic theory and practice. The use of such methods raises the possibility that the Communist party might eventually lose its control over the economythat the formulation of economic objectives will be taken over by technicians. 10 Some specific economic applications of cybernetics in the Soviet Union are: • Systematic control of planning and administration of large construction enterprises. • Selection by computer of the optimum method of transporting a raw material between its producers and its consumers. • Simulation of production processes in machine and tool design. • Development of a city master plan which even specifies the optimum time period for urban renewal. In fact, the control of the total national development is considered a part of cybernetics. In the legal field, the Soviet objective is to use cybernetics to provide a scientific basis for law and law enforcement. Numerous Soviet writers have suggested that cybernetics can be used to codify the laws and to analyze their internal consistency. Although the application of cybernetics to legal problems looks promising, few, if any, cybernetic techniques have yet been used in ~ctual practice. The major impediment is the technical problem of converting legal information into computer language. Some work has started recently on collecting and analyzing legal statistical data in an effort to identify conditions that produce crime. The aim is to find means for detecting criminal tendencies at an early stage. Considerable effort is also being made to automate the analysis of fingerprints, handwriting, facial features, and of substances such as metal shot, glass, and paint. In the field of medicine, Soviet practitioners look to cybernetics to convert medical practice from an essentially qualitative art to a quantitative science. They say man can be regarded as a cybernetic system that is subject to the same COMPUTERS and AUTOMATION for April, 1967 general laws of control and communication as inanimate systems. A great deal of effort has been directed toward automating medical diagnosis. Several medical centers are reported to be using diagnostic machines. No remarkable results have yet been reported in literature on these developments, nor on the development of automatic selection of treatment. Cybernetics is also being applied in the field of public health, with emphasis on the collection and analysis of medical data. Some interesting work is being done in establishing optimum diets and hygienic norms of work and rest. In the field of education, particularly in the area of pro- grammed instruction, the Soviets have borrowed extensively from Western methodology and textbooks. Interest in programmed instruction is high, and a rapid increase in its use- is anticipated. Teaching machines are under active development, particularly in the military and higher academic institutions. In general, there are fundamental differences between the U.S. and U.S.S.R. in the development and application of cybernetics. The environment in the U.S. appears to be more conducive to creative breakthroughs in cybernetic theory, while the Soviet environment appears to be more conducive to large-scale application of cybernetics. COMPUTERS CHECKING ALL UNITED STATES TAXPAYERS FOR FIRST TIME Norman P. Teich Honeywell Electronic Data Processing Wellesley Hills, Mass. 02181 Computers are checking all Federal income tax· returns for the first time this year; not some them, but all them. By mid-April, filing deadline for most taxpayers, seventeen Honeywell 200 computers in the seven Internal Revenue Service (IRS) regional service centers will have received for processing more than 100 million tax returns representing more than $130 billion in collections. According to IRS Commissioner' Sheldon S. Cohen, all returns are processed in the service centers before being sent to the National IRS Computer Center in Martinsburg, West Virginia. At the National Center, the returns are matched against the master files, which include a record for every single tax paying entity in the country. The files bring -together in one place for thorough scrutiny all tax transactions of all taxpayers - no matter where they live or when they filed their returns. Every taxpayer has only one account and all information affecting his tax status is matched against that account and no other. This nationwide system was developed during the past five years. Its need became obvious when IRS projected its workload and realized it would become virtually impossible to collect and account for tax revenues - while effectively enforcing tax laws - unless a significantly different approach for administering the system was developed. The magnitude of the job to be accomplished can be realized from a few simple statistics: in 1930, only six million tax returns representing $3 billion in revenue were filed; this year, more than 100 million returns will be filed and $130 billion collected. Tax returns are "perfected" at each of the seven regional centers by tax examiners whose job it is to make sure all information asked for is entered on the form in the correct places. The information IS then transcribed onto punched cards and fed into the Honeywell 200's, which verify the returns' mathematical accuracy and make some preliminary at at calculations on deductions, exemptions, etc. Magnetic tape records of the returns are then prepared and sent to the National Computer Center, where they are matched against the national master files. Each taxpayer's account includes the taxpayer's permanent tax number, a continuously updated multiple-year digest of tax data about the tax account, data on all returns for which the taxpayer is liable, when and where returns have been filed, the amount and status of each liability and any audit results. The National Computer Center, when it has completed its job, produces a number of magnetic tape reports which disclose who failed to file returns, owes taxes for previous years, has refunds coming or filed duplicate claims for refunds. Also revealed are any discrepancies or unusual characteristics of a return that would suggest a need to examine it further. Refund tapes listing amounts to be returned to taxpayers are sent to U.S. Treasury Department disbursing offices where refund checks are printed and mailed. Other tapes are sent to the regional service centers, where they are used to prepare tax bills, delinquency notices and other communications to taxpayers, and identify returns that are to be considered for audit. IRS' nationwide system - now complete - provides for the first time a comprehensive, uniform, nationwide check on failures of individuals or businesses to file returns. It also improves mathematical verification of returns, increases the ability to detect improper refurid claims, and enables more complete checking of data reported on information documents - such as dividend and interest payments - against information on tax returns. In addition it will insure proper application of credits and reduce processing errors. The capacity and memory of the system benefits taxpayers who may have overlooked prior year's credits or advantages provided by law and enables credit of undelivered refund checks to tax accounts. IFIP ADMITS HUNGARY AS 26th MEMBER COUNTRY International Federation for Information Processing IFIP Congress Office 23 Dorset Square London, N.W.I., England The total membership of the International Federation for Information Processing was brought to 26 nations with the formal admission of Hungary at the eleventh General Assembly of IFIP. Hungary is the fifth eastern European country to join the Federation. The other four are U.S.S.R., Poland, Czechoslovakia, and Bulgaria. COMPUTERS and AUTOMATION for April, 1967 At the meeting, both Yugoslavia, whose admission is under consideration, and Mexico issued invitations for IFIP Congress 71 to be held in their respective countries. As previously announced, IFIP Congress 68 will be -held in Edinburgh, Scotland, August 5-10, 1968. 11 "INDIVIDUAL PRIVACY AND CENTRAL COMPUTERIZED FILES" - COMMENTS I. From Sherman C. Blumenthal Sr. Systems Planner Union Carbide Corp. New York, N.Y. 10017 The following letter was sent to Senator R. F. Kennedy, Senator J. Javits, and Congressman Leonard Farbstein, relating to your editorial in the October 1966 issue of Computers and Automation: The invasion of privacy of the individual citizen by various government agencies is' an increasingly disturbing phenomenon to many thoughtful people. Deliberate and widespread patterns of government snooping of questiomible legality have come to light. This includes such reprehensible practices as the monitoring of first class mail in the Post Office, unauthorized wiretaps by the Federal Bureau of Investigation and the Internal Revenue Service, and a suspicion of ubiquitous, rampant and virtually undetectable electronic eavesdropping. Most disturbing of all, perhaps, are recent proposals to establish a 'centralized computer data bank, collecting all the information about each individual available throughout the government in a single file. The existence of this capability will whip imaginative bureaucrats into a frenzy of unrestrained abuse. The temptations are simply too great for us to believe that the desire for information will not outpace the ability to restrict its use. I have enclosed what I believe to be an excellent editorial on this subject from a recent issue of Computers and Automation. As a practicing professional in the computer sciences I can vouch for the reality of this threat to'individual freedom. This development should be stamped out now, firmly and forever, by enacting legislation along the lines suggested in the ed~torial. II. From the Editor Thank you for the copy of your letter to the Senators and Congressman. Weare pleased that you thought the editorial worth sending on to them.' "BOOTSTRAPPING A CAREER IN THE COMPUTER FlELD" - COMMENTS I. Fred V. Rubenstein Los Angeles, Calif. As an avid reader of Computers and Automation I have been active in promoting many of the philosophies you touch upon in your C&A editorials. If possible, I would very much appreciate receiving ten copies of your September 1966 editorial entitled "Bootstrapping a Career in the Computer Field." I have shown my copy (still intact in my personal file) to several of my associates who enjoyed your message. How nice it would be to circulate copies among some of our high school and college students. Some of us are interested in contributing articles to your magazine. Please indicate what the appropriate steps would be to submit an article to Computers and Automation. 12 II. From the Editor We appreciate your kind remarks, and we are happy to provide several copies of the September 1966 editorial. If you want still more, we suggest you make arrangements with some local copying service. We would like to have an article from you; some information for prospective authors is enclosed (see the notice "Articles Jor Computers and Automation in this issue). ECPI PROGRAM WRITE-UPS AVAILABLE A. Estipona, Technical Director Electronic Computer Programming Institute, Inc. Empire State Bldg. New York, N.Y. 10001 In order to facilitate testing of 1401 and 360 programs . written by our students, Electronic Computer Programming Institute (ECPI) has developed two resident core dump programs for its IBM 360 16K model 30. It has occurred to us that these programs may have a wider appeal and we are therefore willing to supply a write-up and an object deck to all requestors. Both programs reset themselves each time they are executed and need only be loaded once. In addition, both programs permit periodic dumps to be taken by inserting the coding given in the write-up at the appropriate point in the user's source program. When this option is used, all registers are restored and control is returned to the user's program. Our 1401 dump program is designed to overcome the problem of obtaining the contents of the read area when taking a storage dump in the Compatibility mode. The program solves this problem by being loaded prior to testing and remaining resident in high core. A storage print, starting at the read area and continuing to 14,999, may then be obtained whenever needed simply by altering to the starting address of the dump program (15,000). The primary restriction on the use of the program is that any programs to be tested must not have a CTL card greater than 55 (12K). This is to pre~ vent the dump program from being cleared when the user's program is loaded. For testing of 360 programs, ECPI uses a supervisor generated by the BPS 8K tape System. To obtain a storage print after a student's program caused a program check interruption, we .used IBM's storage print program. However, since this program included its own supervisor, it was neces,sary to re-Ioad our supervisor before we could resume testing. To avoid this extra step, we developed a resident storage print program which is placed immediat~y behind the super7 visor and loaded with it. The program modifies the instruction address in the new Program Check PSW so that linkage is automatically provided to the storage print program in the event oJ a program check. The contents of the sixteen registers, the interruption code, the address of the next sequential instruction, and -the contents of main storage are printed in hexadecimal format on SYSLST. Control is then returned to the program loader in the supervisor.. The two restrictions on the use of the program loader in the supervisor must not go beyond address 2999 and programs to be tested must not begin before address 4000. The program as assembled, will print the contents of storage beginning at address 0000 and ending at address 16,000. As an additional feature, both prognims can be modified to print fewer positions of storage by modifying the beginning and ending address constants. Additional information on this aspect is provided in the program write-ups. The write-ups and object decks rilay be obtained by writing to me. COMPUTERS and AUTOMATION for April, 1967 JOVIAL BULLETIN ESTABLISHED - PAPERS SOUGHT David K. Oppenheim Abacus Programming 3507 Barry Ave. Los Angeles, Calif. 90066 A new information bulletin has been established to deal with the Jovial programming language. The new Jovial Information Bulletin (JIB) will be published under the auspices of the Special Interest Committee on Programming Languages of the Association for Computing Machinery, and will be distributed (along with the Algol Bulletin, the PL/I Bulletin, and the Cobol Information Bulletin) as a supplement to SICPLAN Notices, the Committee's informal monthly newsletter. As editor of "The JIB," I am seeking contributions concerning Jovial that may be of interest to Jovial users and implementers, as well as to computer programmers in general. Suggested topics include: • Comparisons of Jovial with other programming languages, outlining advantages, disadvantages, etc. • Interesting applications of Jovial, including real-time and non-real-time applications. • Language features that are compiler or machine dependent and are not generally available in Jovial systems. • Information about Jovial processors, including machines involved, current status, language restrictions or extensions, interesting processing techniques or outputs, and error checking capabilities or incapabilities. • Algorithms for processing Jovial source-language programs, as well as algorithms in Jovial for which the language is especially applicable. • Philosophy and techniques of Jovial programming, including coding, debugging, and operation, machineindep~ndent programming methods, use of and extensions to the com pool, etc. . • Improvements, extensions, and deletions that would be desirable in Jovial. In addition to contributions on these and other topics, I would like information from: those in charge of programming groups or computing facilities that use Jovial as a programming or documentation language; those with development or primary maintenance responsibility for an existing, developmental, or planned Jovial compiler or processor; those interested in receiving copies of "The Jovial Information Bulletin." Please address all comments and contributions to me at the above address. (Please turn to' FORUM,. page 53 ) c&a PROBLEM CORNER Walter Penney, C.D.P. Problem Editor, Computers and Automation Readers are invited to submit problems (and their solutions) for this column to: Problem Editor, Computers and Automation, 815 Washington St., Newtonville, Mass. 02160. This month's problem: Claude Liffey picked up a piece of paper from one of the desks and was studying it when John Lawthorne came in. "Did you forget we were going to have lunch together today?," his fellow-professor asked. "Oh, I'm sorry. One of the students in my programming class left this behind and I was trying to figure it out. Must have gotten a little too engrossed." He handed the sheet to John (see Figure 1 ) . "I was going to punch up a few cards and run the program just to see what would come out," said Claude. He watched as his companion hit a few keys on the desk calculator, then wrote a number at the bottom of the sheet. "No need to do that. Here's the value of z that would have been printed out," he announced. What did he write? COMPUTERS and AUTOMATION for April, 1967 Solution to last month's problem: The number waS 1 001 1 1 0. Whether interpreted as a decimal or binary number (i.e., 610), this leaves the same remainders when divided by 3, 5 and 7. But when divided by 11, only 1,001,100,010 (decimal) leaves a remainder of 9. °°° 13 c&a MARKET REPORT TOP 100 COMPUTER USERS AMONG INDUSlRIAL FIRMS HAD OVER $ 2 BILLION WORTH OF COl\1PUTERS AT WORK BY MID1966 Over a third of the market for computers among industrial firms consists of the one hundred top accounts. These hundred (the "EDP 100") have over $ 2 billion worth of computers installed, and they account for 34.~/o of the total value of computers used by industrial corporations. Indeed, these hundred companies use nearly 16% of the $13 billi'on worth of computers currently installed in the world (not counting military applications which are classified) • These figures are derived from a study of the value of computer and related data processing equipment in use by the 500 largest industrial corporations in the U.S. and the 200 largest foreignbased industrial concerns, as compiled by Fortune magazine, July and August, 1966. The study was conducted by International Data Corporation. The top 100 computer customers -- the "EDP 100" -- are listed lJelow. is virtually no way to separate computers used for computer product development, programming, a~d sales support, from those actually used for lnternal business in such firms. Interpretation The world's largest manufacturing corporation, General Motors, heads the list as the world's largest computer user among manufacturers. It was using $175 million of computers as of June 30, 1966, the cutoff date for the study. Next down the list was General Electric, with $85 million of computers, followed by Boeing with $80 million, Lockheed with $75 million, and Ford with $72 million. Other firms tend to line up by size-within their industry classifications, but there are quite distinct characteristics of computer usage among different industry groups. As Table I indicates, 12 aerospace firms on the EDP 100 list account for 21.8% of the computers included in this study. Preparing the Base To qualify as "industrial corporations", and thus be included in this study, firms must have received at least 50% of their revenues from manufacturing or mining during the most recent fiscal year. American Telephone and Telegraph, a utility, would therefore not qualify, even though it is probably the largest computer user outside the Federal Government. However, Western Electric, its manufacturing affiliate, is included in the EDP 100. Also, companies whose primary products are digital computer systems -- such as IBM and Control Data -- were not considered for the EDP 100. There TABLE 1 EDP 100 Foreign Use Although 35 of the EDP 100 are foreign-based, they use only 18.~/o of the computers. To some extent, this indicates relative immaturity of the foreign marketj one of the reasons for the higher growth rate currently being achieved 'in this market is that foreign firms are catching up with their U.S.-based firms in the value of computers used to support their business activities. For more information about complete IDC survey results, designate 4 on the Readers Service Card. DISTRIBUTION OF VALUE OF COMPUTERS WITHIN INDUSTRY GROUP ($ Million) Aerospace Automotive Chemicals Electrical Food & Beverage Petroleum Primary Metals Rubber Miscellaneous $450 14 ,COMPUTERS and AUTOMATION for April, 1967 a· ''Alii know is one evening in our living room, my wife and I decided it was agood idea to jOin the Peace Corps. So we did:' What the David Kadanes did puzzled and puzzles a lot of people. Maybe because tho Kadanes weren't anywhere near twenty years old anymore. Maybe because they gave up two years' worth of a lot of salary, two years' worth of a big job as General Counsel for the Long Island Lighting Company, two years' worth of a life they had spent their whole lives building ... just to join the Peace Corps. But what a lot of people don't realize is that the Peace Corps isn't just a place for justout-of-college kids with strong arms and heads and good-size hearts. The Peace Corps is a place for p·eople who want to do something and can do something. It's a. place for people who want to see things and do things firsthand and closeup. People who want to give other people a chance to get to know and under· stand their country and themselves as they really are. People who care about the world and other people maybe even as much as they care about themselves. And, maybe more than anything, the Peace Corps is a place for people who, for some reason, are willing to give up whatever they have to give up to do something they.feel they have to do. And the David Kadanes are two of those people. Write: The Peace Corps, Washington, D.C. 20525. ..... :\\ ~ f~'i 1-~; PUBLISHED AS A PUBLIC SERVICE IN COOPERATION WITH THE ADVERTISING COUNCIL COMPUTERS and AUTOMATION for April, 1967 15 REAL-TIME PROCESSING POWER: A STANDARDIZED EVALUATION J. Burt Totaro~ Associate Editor Auerbach Standard EDP Reports Auerbach Corporation Philadelphia~ Pa. 19107 ((Most current computer systems offer several different types of randomaccess storage devices, with speed and storage capacities (and prices) that vary by several orders of magnitude. H ow can the user determine what computer system can best serve his growing real-time needs?" A clear trend in the computer industry today is toward the use of computer systems that employ large-capacity randomaccess storage devices and perform "real-time" processing in a wide variety of application areas. Real-time processing in its broadest sense can include any type of processing that for practical purposes is instantly responsive to external events and is capable of influencing the further course of these events. Thus, both missile guidance and simple on-line file inquiry systems can be considered rcal-time activities. Applications Random-access devices and the real-time processing they permit have become so important that many current computer systems require some kind of mass storage· device in every equipment configuration that is sold. These devices can aid considerably in improving the performance efficiency of a system's hardware and software. Fast-access storage devices· make possible efficient real-time processing in such diverse applications as: • • • • • • • Inventory control Airline reservations On-line savings Process control Credit checking On-demand management reporting Conversational programming. - The performance advantages and application flexibility gained with real-time processing in random-access-oriented systems have forced almost every current or prospective computer user to come to grips with questions such as "Are we ready for real-time processing?," or "What computer system can best serve our growing real-time needs?" The user contemplating installation of real-time hardware or addition of another real-time application to his -system cannot easily find answers to his questions because the subject matter is complex and the consequences of any decisions are far-reaching. Evaluating, selecting, and utilizing real-time equipment is especially difficult because up to this time there has been no effective method - short of full-scale (and expensive) simulation - of measuring the performance of planned real-time "ystems. New Estimating Technique This article discusses a new standardized estimating procedure for measuring the ability of computer systems to locate and update randomly addressed records in real-time applications. It is based on a standard random-access benchmark problem, which is similar in concept to the standard sequential benchmarks. 1 Using a detailed programming technique that is standard, objective, and yet flexible, the new estimating procedure measures computer systems in their performance of the random-access benchmark problem. The resulting performance times serve as a valuable guide to the overall performance of those systems which are oriented toward the use of random-access devices. They also serve as a useful tool in comparative system evaluations, since all systems can and will be measured in basically the same standard manner. Real-Time Benchmark Approach One of the most common real-time jobs in commercial data processing installations is the processing of randomly (and often remotely) entered detail transactions against a master file stored in random-access storage devices. The detail transactions supply the information used to update the master file by inserting new records, deleting old records, and modifying 1 "Standardized Benchmark Problems Measure Computer Performance," John R. Hillegass, Auerbach Corporation, Computers and Automation, January 1966, pages 16-19. COMPUTERS and AUTOMATION for April, 1967 the contents of existing records. Usually an activity report is prepared concurrently to reflect the modifications made to the on-line master file. Figure 1 illustrates this type of application. An on-line savings system is a typical exarriple of this for.m of real-time processing. Computer F'igure 1. Rllil Diagram of Random-Access Benchmark Problem The application parameters that ~ave the greatest effect on throughput ,times in real-time processing are: organization of the master file in random-access storage, source and frequency 'of real-time detail transactions, amount of computation required to update each master-file record, and method of handling the activity report. With regard to these parameters, the random-access benchmark problem (like any standard benchmark problem) includes certain necessary assumptions and parameter specifications. The following paragraphs explain these basic assumptions and general parameters. Master File Organization The master file is assumed to be stored on one or more random-access storage devices in sequential order by item number. It is further assumed that the master records will be accessed through use of the indexed s2quential access techriique, with a complete cylinder index permanently resident in core storage and a track index stored on the first track of every logical cylinder. 2 Thus, when a detail transaction enters the system, the cylinder index in core storage is searched to find the number of the cylinder that contains the desired master-file record. Seek Operation • A seek operation is then performed on the random-access device to locate the desired cylinder. Next, the first track of the cylinder is read to obtain the track index, which reveals the exact track and block number within which the desired 'record is stored. Using this information, the desired master record is read into core storage for updating. As mentioned;' the master file is assumed to be distributed in strict sequential order through the random-access storage medium. One reason for this assumption is to make the storage of the master file as application-independent as possible and thereby extend the general applicability of the results. It is fu'rther assumed that there are no known differences in the activity (i.e., frequency of access) of individual master-file records, so that no performance advantage can be gained by organizing all highly active records in the same or consecutive cylinders. Experience has shown that the master files in 2 A "cylinder" is that group of tracks or data storage locations that can be accessed when the movable access mechanism of a random-access storage device is positionecl at one discrete position. A fixed-head drum unit or disc file would, therefore, consist of only one actual cylinder since all of the data it holds can be accessed without repositioning of the access mechanism. In practice, however, the cylinder concept. can be - and usually is - applied to fixed-head devices by subdividing the total body of data tracks into a number of logical "cylinders" so that the. two-level indexing scheme, as used in this benchmark problem, can conveniently be applied. COMPUTERS and AUTOMATION for April, 1967 many application areas are organized in just this straightforward fashion, especially in their earliest versions of implementation. Detail Transactions -The random-access benchmark problem primarily measures the speed at which a given system can locate, read, update, and write randomly speG,ified master records on randomaccess storage devices. In order to ensure that the benchmark will accurately measure the throughput capabilities of random-access devices engaged in file-updating activities (and . to ensure that the times for this primary activity will not be muddled by extraneous and widely varying considerations), no attempt' has been made to include timing considerations involved in obtaining the Detail Transadions. Instead, the Detail Transactions are assumed to be arriving in random sequence and at a continuous rate from either local or remote input devices. Wh-en the transactions enter the central computer system, they are stored in a transaction queue in a communication area of core storage by a r~al-time program or simple data input routine resident in a "foreground" portion of core s'torage. The random-access benchmark problem, then, obtains its Detail Transactions by accessing the common communication area of core storage. It is assumed that the Detail Transactions arrive at a sufficiently high rate so that one or more transactions are always waiting to be processed; i.e., the transaction arrival rate is nev~r allowed to be a limiting factor upon the computer system's throughput. Through this assumption, the applicability of the results of the standard benchmark test is again kept broad. To the random-access benchmark problem, it makes no difference whether the Detail Trqnsactions enter the system via a local card reader, via a simple remote inquiry device, or via a complex data communications network. Detail Transactions are considered to be device-independent. The random-access benchmark problem primarily measures a system's ability to locate and update randomly addressed master records. It does not attempt to measure the efficiency of the system's data communications network. Required Computation The amount of computation required to update each master-file record is determined by the number of computatiRnal steps required to perform the program functions specified in' a series of detailed flowcharts. The general flowchart in Figure 2 summarizes these detailed flowcharts and indicates the basic computational processes that must be performed. Because of the imbalance between computational speeds and the access times for most current random-access devices, the central processor time required to perform the spe~ified computations will generally be a small portion of the total time required to update one randomly addressed master-file record. The amount of computational time can, however, affect the number of disc or drum revolutions required to complete the processing cycle for a single master record. The central processing time for each system is clearly indicated in a Worksheet Data Table that lists both processor and input-output times required to perform each logical function of the standard problem. Standard throughput pertormance graphs show both the elapsed job time and the total amount of central processor time used in processing the master records. Comparing elapsed and central processor time will show the amount of central processor time available for use by other jobs. This information is of vital interest when measuring computer systems that are ~capable of multiprogrammed operation~. These graphs and worksheets form part 17 sequential access technique to obtain Master Record Address Search cylinder index in core storage for address of cylinder holding Master Record; read first track of cylinder (Track Index) for location of Master Record in cylinder. Read track or portion of track containing Master Record and isolate Master Record. Using information from Detail Transaction in buffer area, update stock control data in Master Record. Move next Detail Transaction from com- access storage for master-file storage; the medium-scale standard configuration must have at least 20,000,000 bytes; and the large-scale standard configuration requires at least 100,000,000 bytes. Most current computer systems offer several different types of random-access storage devices, including magnetic drum, magnetic disc, and magnetic card or strip devices. The speed and storage capacities of these devices (and their prices) can vary by several orders of magnitude. The principal guideline that is followed in choosing the particular random-access storage devices to satisfy the auxiliary storage requirements of each standard configuration is economy, both in terms of the monthly rental price per unit of storage and the throughput potential of the selected devices. Other guidelines are that the selected device must be fully software supported and must have reasonably high reFability. munications region to buffer work area. Benchmark Description Put Master Record back in its original location Write updated Master Record and re-read to verify recording accuracy. Format Activity Record and write it on tape, disc, or printer Compose line of print for Activity Record and write to fastest available peripheral device for later off-line printing. Figure 2. General Flowchart for Random-Access Benchmark Problem of the standard method of presenting the results for all realtime systems measured. Activity Re~ort The standard format of the Activity Report record is rigidly specified. This record of master-file activity will generally be written either on a magnetic tape unit or on a random-access device that is not used for master~file storage. The Activity Report itself is assumed to be printed off-line at some later time. (If a system does not have the capability to use magnetic tape units and cannot add additional randomaccess devices, the Activity Report record will be printed on-line, though this mode of operation may severely reduce the system's throughput.) It is further assumed that the Activity Report record will be made available in core storage for access by a separate routine for possible transfer back to the remote terminal that entered the Detail Transaction (as in most inquiry jresponse applications). Decisions concerning the use of either magnetic tape or random-access devices for the Activity Report records, and amount (if any) of record blocking, are made in such a way as to achieve the best possible throughput rates. Standard Configurations Real-time performance estimates derived from the randomaccess benchmark problem are meaningful only in relation to the hardware equipment configuration that "performs" the benchmark. Therefore, three standard system configurations have been specified, and each computer system's performance is evaluated in one or more of these standard configurations. Standard configurations permit valuable performance and price comparisons to be made between different computer systems, and also permit computer users to evaluate performance results in relation to the standard configuration that most resembles their current or proposed random-access system. The small-scale standard configuration requires (among other devices) at least 5,000,000 bytes of on-line random- 18 Like the widely-utilized generalized sequential file processing problems, the random-access benchmark problem represents a typical iriventory control application. The basic processing performed in an on-line inventory control application will closely resemble the basic processing performed in many other real-time processing applications. The sequential file processing problems generally use master files stored sequentially on magnetic tape, accept sorted Detail Transaction records (usually from punched cards), and print the Activity Report on an on-line printer. The random-access benchmark problem, by contrast, uses master files stored on random-access devices, accepts Detail Transactions in random order, and writes the Activity Report on auxiliary storage devices. Aside from these basic differences in the method of sorting and handling the files, the sequential file processing problems and the random-access benchmark problem share many characteristics, including the entire computational procedure that updates the active master-file record. Thus, valuable cost and throughput comparisons can be made between sequential and random processing techniques. Format The basic form of the random-access benchmark problem is represented in Figure 1. Record layouts for all files are specified in detail, but the format and packing of the masterfile records can be slightly modified to permit more effective utilization of the hardware being measured. The detailed processing steps in the random-access benchmark problem are clearly specified by' flowchart diagrams (as summarized in Figure 2), but the exact programming procedures used to code these processing steps are left to the analyst performing the estimate. Thus, the analyst can take full advantage of the hardware features and peculiarities (such as variable word length, special arithmetic capabilities, input-output overlaps, special random-access instructions, etc.) of the system he evaluates. This technique adds to the validity of the resulting performance measurements, since the hardware is always used to its best advantage, just as programmers strive to use it in actual applications. Automated estimating procedures, by contrast, tend to be overly general and can be hard-pressed to supply custom-tailored measurements that take into account each system's unique characteristics. Presenting the Results The estimating process is separated into a series of logical steps, the results of which are summarized in the Worksheet Data Table (Figure 3) that is prepared for each computer COMPUTERS and AUTOMATION for April, 1967 .. WORKSHEET DATA TABLE 2 - RANDOM ACCESS BENCHMARK PROBLEM CONFIGURATION TASK IVR Flowchart Block 1 E.1- E.4 E.5 E.6 - E.8 Random Access Benchmark Problem Timings, msec E.9 - E.12 E.13 - E.22 E.23 - E.30 E.31 - E.34 E.35 - E.39 1 'unction Central Processor Time, msec Housekeeping Seek and Read Track Index Search Track Index Get Master Record Update Master Record with Transaction Get next Transaction Write updated Master Record Format and write Activity Record 1. 51 0.76 Standard resident routines Fixed overhead Benchmark program File and work areas Cylinder Index Track Index 3 Random Access Benchmark Problem File Characteristics Master Record size, bytes Block size, bytes Blocks per track Tracks per Cylinder for Master File storage Cylinders for Master File storage Devices for Master File storage 1. 52 3.35 41.3 I 0.15 49.1 0.20 17.5 0.70 17.5 0.99 0.30 0.30 0.51 4:200.'712 0.51 50.0 0.95 50.0 1. 56 0.91 Total Time per Master Record, msec 2 Random Access Benchmark Problem Core Storage Space, Bytes VIIIR REFERENCE Dominant I/O Dominant I/O Device/Channel Time: Central Processor Device/Channel Time: Time, msec Master File, msec Master File, msec 0.91 108.8 5.79 116.6 9.34 8,000 128 4,740 1,712 1,890 1,40r. 17,000 128 4,740 2,608 1,985 6,300 Total Core Space, Bytes 4:200.715 32,761 Figure 3. 88 616 5 8 88 1,260 5 18 378 397 3 4 4:200.716 Example of a Worksheet for a Third Generation Computer © 1966 AUERBACH Corporation and AUERBACH Info, Inc .• system. Each logical step (or program segment) is keyed to a series of flowchart blocks that describe' the step in detail. Central processor and input-output times are always listed separately in the Worksheet. Separate times are listed for each system configuration measured in its performance of the random-access benchmark problem. Input-output time is represented on the Worksheet by a single column of figures for each configuration measured. This column shows the dominant I/O device time or the dominant I/O channel time, whichever is greater. Generally, this column of figures, Dominant I/O, will reflect the master record physical .I/O time, i.e., the time required to locate, read, write, and write-check one master-file record (with the time required to perform each operation specified individually). Writing of the Activity Report record will usually be completely masked behind the dominant master record I/O operations. Estimated Total Time The Worksheet Data Table also includes a figure called "Total Time per Master Record, msec" for both the Central Processor and Dominant I / a columns. The higher of these two total figures represents the estimated total time, expressed in milliseconds, to process one randomly addressed master-file record according to the specifications of the random access benchmark problem. COMPUTERS and AUTOMATION for April, 1967 The central processing and elapsed I/O timing figures presented in the Worksheet Data Table are summarized and extended on system performance graphs which indicate the time, in minutes, required to update from 100 to 100,000 master-file records. Applying the Results Since the results of the random-access benchmark problem are presented in logical timing "kernels" keyed to logical sections of the standard problem flowchart, a systems analyst can easily obtain performance estimates for his specific realtime processing jobs by substituting or adding his estimated timing kernels for program segments that are unique to his problem programs. Thus, he can arrive at a new total time (central processor and I/O) to process one randomly-accessed master record in his specific program. This new real-time performance estimating technique has already been successfully applied to a number of thirdgeneration real-time systems, including six models of the IBM System/360. In applying the real-time measures to current systems, widely diverse random-access devices have been used for master-file storage, providing a good test of the flexibility of the technique. The estimating procedure' has shown itself to be objective, flexible, easy to use, easy to modify, and capable of producing valuable measurements of real-time processing power., 19 13th ANNUAL EDITION OF THE COMPUTER DIRECTORY AND BUYERS' GUIDE the regular June issue of carn"'uters r:'nd automation to be published in June, 1967 Contents • A Roster of Organizations in the Electronic Computing and Data Processing Industry. • A Buyers I Guide in the Electronic Computing and Data Processing Field. • A Special Roster of organizations supplying entire equipments for computing and data processing systems. • A Special Roster of computing, data processing, and consulting services. • A Special Roster of commercial time-shared computing services. • A Special Roster of commercial organizations supplying courses, training, or instruction in computing, programming, or systems . • A Special Roster of organizations specializing in leasing or financial services in the computer field. • Characteristics of General Purpose Digital Computers. • Characteristics of General Purpose Analog Computers • A Roster of School, College and University Computer Centers. • A Roster of Computer Associations. • A Roster of Computer Users Groups. • A List of Over 1100 Applications of Electronic Computing and Data Processing Equipment. Price • $12. 00 separately • Prepublication price for Subscribers to "Computers and Automation, " whose present subscription does not include the "Computer Directory" (magazine address label is marked *N) . . $7 .. 50 • Prepublication price for Nonsubscribers . . . . . . . . . . . . . . . . . . . . . . . . . . $9.95 (Notice: If address label is marked *D, your subscription includes the "Computer Directory") SEND PREPAID ORDERS TO: Designate No. 18 on Reader Service COMPUTERS AND A UTOMA TION, 815 Washington Street, Newtonville, Mass. 20 Card 02160 COMPUTERS and AUTOMATION for April, 1967 Small computers are Innovation machines The power under the direct control of the scientific investigator with small computers is enormous. Not just the power to calculate. Not just the power to analyze. But the power to innovate. Small computers - and by that we mean laboratory , size or smaller general purpose machines - are most frequently used on-line, in real time, intimately hooked up to the experiment. Inputs to the computer are direct from the experiment. Outputs are instantaneous detail, or instantaneous sampling, or instantaneous averaged results, all provided while the experiment proceeds. Two things can now happen. First, the investigator can change the method of data taking, or the sequence of data taking, based on the results he sees emerging, He can sample more frequently than he thought necessary, or can look in detail, at an unexpected result. Second, he can influence the experiment, either by himself or automatically by the computer, based on the results he sees emerging. If he waited to complete the experiment and analyze the results at the computation center, the opportunity . to innovate would be gone. Small, real-time, on-line, general purpose computers are intensely personal, highly approachable, responsive scientific innovation machines. It will be the rare laboratory that will not have one in the next few years. DIGITAL is the largest supplier of small computers to the scientific disciplines, PDP8/S, PDp·8, L1NC-8, PDP-9, PDP·9 Multianalyzer, and five versions of the PDP-10, Write for the just-published 540-page Small Computer Handbook and Primer. Free. t COMPUTERS· MODULES DIGITAL EQUIPMENT CORPORATION. Maynard, Massachusr:tls 01754. Telephone: (617) 897·8821 • Cambridge, Mass, • New Haven. Washington, D. C.• Parsippany, N. J•• Rochester, N, Y .• Philadelphia. Huntsville. Pittsburgh. Chicago. Denver. Ann Arbor. Houston· Los Angeles. Palo Alto· Seattle. Carleton Place and Toronto, Ont. • Reading, England • Paris, France • Munich and Cologne, Germany • Sydney and West Perth, Australia • Modules distributed also through Allied Radio Designate No. 11 on Reader Service Card COMPUTER PERFORMANCE PROJE,CTED THROUGH SIMULATION Fred C. Ihrer Vice President and Technical Director Comress Washington, D.C. 20018 "The implementation of a data processing system is as complex as any function faced by management - and evaluation of computer hardware can no longer be realistically accomplished simply by comparing hardware performance data." Increasingly, computer management is recognizing the need to employ scientific_methods to assist in the complex tasks of hardware and software selection and evaluation. Prominent among the scientific techniques available today is an integrated series of computer programs called SCER T (Systems and Computer Evaluation and Review Technique). The SCERT Program has been used extensively for over four years by diversified groups of commercial and governmen"t organizations throughout the United States, Canada and Europe. Proper management of the total environment in data processing requires realistic prediction of computer performance at a number of different points in the total cycle of selection and implementation. Feasibility Analysis Today's computers have the inherent capability to do just about anything from playing chess to completely controlling complex manufacturing operations; but this capability does not necessarily prove feasibility. The planning for new computer applications should include precise measurements both of the cost of implementing and processing, and the response that the computer hardware and system design can provide. the industry for any length of time have seen the explosion of implementation costs caused by the selection of inadequate hardware. Frequently, a bad computer selection requires complete restarting of the whole implementation process. Just as important is the delay of important missions and programs. Far less frequently, improper selection can result in too much hardware power with obvious excess cost. The ability to accurately predict performance when selecting is a vital need in most organizations. Hardware and software systems now offer different "approaches to the solution of anyone problem on a computer. Measuring the performance of alternative designs in terms of cost and response can lead to an optimum design selection. Program Design As computer hardware has become more reliable, the capacities available to programmers have expanded. A predictive tool was needed to allow the programmer to optimize the capability and capacity utilization of the programs. Just as important is the proper selection of programming language from the several now available for a typical computer system. EDP Management Computer Selection Proper computer selection is a need easily recognized by all levels of computer management. Those associated with 22 Finally, computer management needs a predictive tool to measure operational and programmer resources. Such a tool should enable them to' measure increasing and/or peak COMPUTERS and AUTOMATION for April, 196i workload before the fact, and to plan for obtaining proper hardware at a realistic time. Why Simulate? The evaluation of computer hardware can no longer be realistically accomplished by comparing instruction execution times, peripheral speeds, memory sizes and other unrelated hardware performance data. Other approaches to performance prediction, such as benchmark processing and instruction mix statistical data, are inadequate for the evaluation of computers performing in a multi-programming mode of operation. Simulation is a scientific tool that provides for the analysis of complex structures, procedures or systems by modeling the systems and examining the impact of alternative decisions based on predictive results. The more complex the simulation, the greater the need for a computerized modeL The planning, installation and implementation of a data processing system is as complex as any function faced by management. The number of decisions and alternatives presented at every stage of planning and implementation makes simulation a useful tool. From early stages, when alternate gross automation plans are presented, to the complex decision for propel~ hardware/software selection, each decision is best made in light of the eventual impact of the alternatives on the final performance of the system. Hardware Independence A second basic premise in the design of SCER T is concerned with the hardware independence of the input specifications. Optimum use of SCER T would allow for the evaluation of a defined system on any conceivable configuration of computer hardware and its related software. SCER T would have the responsibility for adjusting the system to conform to the particular hardware being simulated. This IS accomplished in two ways: 1. Factor Library An important part of the SCER T Program is the Factor Library. This Library represents a technical data base which contains the significant electronic and mechanical performance factors for virtually all commercially available computer hardware. The Library also contains performance characteristics of most of the software packages available. This software includes data on compilers, operating systems, Input/Output Control System packages, and sort-merge packages. SCER T uses the Factor Library based on component model numbers defined at simulation time, and constructs a model of the hardware to be simulated, parametered by the specified software. SCERT then produces its first output report which shows general cost and environmental characteristics of the hardware, and serves to confirm the actual configuration being simulated. . The Design of SCERT ~CE~T There are two basic premises underlying the design of SCER T. To properly evaluate the relative ability of computers to process projected (or actual) applications, the total workload to be placed on the computer hardware must be considered. This first premise is particularly true when processing employs on-line real-time concepts or when the hardware and software under consideration is operated in a multi-programmed mode. To provide for the simulation of the total processing workload, the primary consideration is to allow for a rapid but accurate definition of processing requirements. Obviously, SCER T was going to be employed in situations where the processing workload only existed at a high level conceptual design. But in other cases, SCER T would be used to study systems where the programming had been accomplished and the systems were already running on computer hardware. The input to SCER T which describes processing therefore must have a broad range of adaptability but, at the same time, have the capability of being prepared in a minimum amount of time with a minimum amount of analytical effort. AIJALvSIS PAkT 1 CU~lPUTEI{ COMPLEMEr,T REPORT FOR SYSTt::M Xt:Lbb COMPUTER CO~IPL~MENT IJUMtlt.fl Hbb ----------------------------~ fNY! RONMENTAI MANUFACTLIRE-I< U\lAN II TY PURCHASE ~lINIMUM REQUIREMElns _ ___ ___LN _. PR I CE MONTHP'L---,,-FL~O!,!O,,-:RI~N!';tG--,C'O'0~OL~I~NG~P0';7W;;-E,,-R$ RENT~L SQ FT BTU/HR KVA SYSTEM MOi.lt::L NUMtlEK -------CE.NTRAL PROCESS1"G Uf,IT COMPUTEt< MAIN-FRAME 1bl·,/3b. 2.~0H 8"b,95~ 17rl4~ !:>71 37,14. 18.0 366,501:> 7,b0~ 4,94~ 180 144 35, BH0 16.0 8.4 66. 96 2,60. 1.2 1.4 3,80~ 1.5 -------MAGI,E_T l~.JA~E;._.uI~~-10 Wr-./3b. 24.23 ILlM/3b. 24.3!:> 2"3,2B~ " CARLi .REAQEK_______________ 3~' 1 ILlM/36. 2540 PRlr;H.R 14.3fH .B" 22,0~a (tf'-1I 3 6i:1 41,2." 90e 55 4,6~i1 RAI'uQ"l.ACCE:.~S _UI;1 TS Ibl~/3b" ,,3.11 9b",r~. 52,6H. 2,25. 1, l~H 75 50 'I, '64~.45b ~4,6~~ 11111 _llJr~t.36il 23J..1Jc HlI8L. ~ II~ I 5r~IEF e0~ 1.5 l ~9'14i1 ~8,~ HiE: PR1C.ll% ANU __f.WLlKQ~.. ~:E.NIAl-£AC.IO~Ok THE F Oht.,9.'tiilLG EGUIPMEN.LT- - - HAV~ BEE" ADUELJ TO THE DEVICES LISTEU AI:JOVE ------------------------21,p 28.41 26411 1 COIJSO! E 1 MAG TAPE CONTROL 2 CHANNELS I j.d~ ~ _1ul1'~](L b9"~A!:> Ibl~J.3bl:1 b9tllA!>-- luM/3bl1 7101 Jot I Iii /361:1 llW,/3b. ItJ"./36~ .. 1 k/52 CQtyTHOI __ L_.D~E.. CONIROL-I 1 ---- SEL CHAW,EL 2~50 H-S 1ST _---1------SE.L.cHAt1N£L-2.l!!:l~ li-=-S'-"2o.uN~D____ SI.IO HEAD/.R lTt: FOI< 24~2 ThE FOLLu'.lNG 5uFTwARE HAS B~EI·J UTILIZED wlTH THE ABOVE COMPUTER CO~IPLEMENT Analytical Language This design goal was met by the development of an analytical language which is structured similar to programming languages, but at a much higher level. The language has three divisions: (1) the environment division, which provides background data on the system to be simulated; (2) the file division which describes the general characteristics of all files involved in the computer processing; and (3) the systems division which specifies the actual processing to be accomplished on the computer. In a typical situation, a trained SCERT Analyst can define a computer system consisting of approximately 100 runs in one week or less. COMPUTERS and AUTOMATION for April, 1967 2. Presimulation Algorithms The second way hardware independence is achieved in SCERT is through presimulation algorithms. These algorithms structure the processing descriptions so that they are compatible with the particular computer being simulated. The major algorithms accomplish the following: a. A computation of internal processing time and memory requirements. As a by-product, a computation of the number of program steps required to perform the internal processing. 25 b. The assignment of all files to available peripheral devices and channels. c. The restructuring of files in terms of record format, optimum file blocking and buffer areas. d. The calculation of throughput timing for all input/ output functions and their memory requirements. e. The calculation of pre- and post-run overhead time such as program insertion, setup, error correction, and re-run. component involved in the real-time processing and shows the statistical probability distribution of queue sizes in the hardware complex. SCERT HEAL TIME. Ar.ALYSls.I'ART 2 FOR SYSTE~l HARD.-ARCiJtiLTiATiofj---'(;-;CM""INCC:U~T~ES;;-')------ XEL.bb COMPUTEK CO~IPLE~'Ef,jT--N-UM8Ef< XXx---PERiOEi:-e:30. 00 MINUTEs - PART 2A ~IAIN CO~;f.oUTER -SYSrEiof ~,E T Pt:RCENT PROtJAtJLt QUEUe; LENGTH EXPECTED WORST UULIlA.T)_O~(_U_!lLlZED Simulation Methodology l l . ~3 The SeER T Program employs three distinct simulation methods. The first of these derives the elapsed running times for each computer run or real-time event. If the system being simulated is only concerned with sequential batch processing in a non-multi-programmed environment, this stage of simulation is the only one employed. This first stage of simulation is called a throughput simulation. SeERT literally explodes the processing requirements into the maximum number of unique throughput possibilities, and then simulates the flow of processing for each unique situation, thereby producing a projection of net running time. The result of the throughput simulation is then reported in a "Detailed System Analysis," which is produced for every computer run simulated. It also contains certain other data useful for the implementation of the system such as optimum file blocking and channel assignments. SCERT DETAIL SYSTEMS AI,AL.vsis FOR RUN NuM~ER FREQUENCY D C20.3 - cO~IPUTE~- CC~IPI..E~iENT--~iuM8Ek- R66 PAKT iltipU'rloliTPUT-Ai'AL..isfsFILE t.U~'I:lEK _ _ _-"'A'"'LL'"----'-TII~ES ARE IN MINUTES I;O.OF. REC()RD SIZE RECORDS NO. 1/0 tJUFFERED Ufl,;UFFERED RE-CJRLJS __ AVG _J'AX~£lLBLO(;~BE~LS MEDIA _ _ _2.T-"IMC-"E~_ _~T-"IMC"E=----_ DATA CHANNELS 3___8'"-•.::-.44"---_ _ _--"-~-'.:.6~2'----_ _-"-5~.6:><;6,--_ 01 9.45 :31.51 02 1l.70 39.~1 03 _______6.!1l.8_~_.96 5.76 2.CJ7 HAI,LlOM ACCESS ;'ODUL.t:S 814 01 02 4.17 4.10 ~i 803 13.90 l.3.08 0.15 0.14 2.a' 2.1~ --~:;~--;;:~~~-----7.-~~:~~~'-----~~~:=~~'---- __ ~02 ______10. 4"-3_----':3"'4'-'..-'-.77'----_ _ _-"-0.'..:.4~7_ _ _~4~.2~6'____ PART2B Cu"MUf.!0'--Jl~9...:.5u7_'J.9...:.5,..4~_ __ TOTAL. Ur,[;UFFEREG II,PUT/(,UTPUT PART ~ mt----------------7-.8-1-- FUi,C 1101' PROGRAM STEPS ARITI1~,lTIC - COMPUTA-f lU;:"LJECI~IOjJ MJD COr.TROL HAI,uLlI,,, II,PUT/UU [PUT ROuTlI;ES II,PUlIOUTPUT AREA vA T A TOTAL Ii,-f-. COMPuTATION MEMORY U~EO 9tlA 90A 9.81 ClH 'J7L Y7L 10.13 (;20 92P Y2P 5 16 2.27 15.90 9.93 5 61 2.32 200.6b 77.39 21.20 1.22 :39.8~ 19.84 10.75 40.00 19.97 11.15 55.71 30.93 14.50 INTERNAL TI ME 192 1.2.0 9~4 4.845 1'0~0 5.422 _ 1.64jl_____ 8.2ee e 28.5.5 3.63 0.44 1.24 5.06 _ _ _ _ _ _ _ _ 0.00 .3.836 C40 TOT AL PKuGHAI'l T 1 "E 980 12.59 0.28 2.17 _____ CS0 ___ 9BQ __911"---------12...2!>_---'b"-'-"-"L7_---'lu.-"6,~"--_--"-'33ll.~2~1L.-~I!..tBLL.61:>-'4L-~9!..L'jj42~_ __ OVERHEAD TIME CO" 2.01 1.05 980 'tao h _____-'-u ___ ..9I1~ 21.24 TOTAL £:'1 IIPSEu T I j\ 1tl3 betl4 COMPUTE BOUND impact on total performance of various levels of operating systems. 6. System Specifications A natural by-product of SCERT is its portrayal of systems processing requirements in a non-hardware form. SOER T is frequently used to simulate a proposed systems design prior to the time of asking computer manufacturers to furnish hardware bids. An 11~~ SCERT SY~1f.M SPECIFICATIONS o.tl5~_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ___ .1 ,I "4.57 .,+.9-;;- r· A."2 A0~:3 tJ.~4 tl9.5 A".3 JPUT FILE DATA ••••••••••••••••••••••••••••••• • GROUP' FILE CHAH/RECOHD F IE~DS/RCD KA TIO FROM RUN VO~UME AVG MAX ALPH NUM AND MODE NUMBER IS) • CC06 ____ b.l".L .2 FOH SYSTEM 11 0456 COMPUTE BOUND NOTHWG IN QUEUE CO~IPUTER All~ AI". _b9.2~ __ ~l~.e .9.9~ COMPUTE BOUND Btl~5 "1 02 NOTHING IN QUEUE COMPARE COMPARE AOD/SUB MOVEMENT MOVEMENT MOvEMENT COMPARE MUL1/OIV COMPARE CCll CCII! CCl" CC09 ccaa CC08 ccaa CC07 cca7 1 cell • 1 CCll • 1 27 EDIT' 5 0 2 4 11 27 27 0 2 2 10 5 5 5 5 5 5 5 10 8 1 2 3 11 11 1 2 2 1 3 1759 cell • 875ccrr-o1 celB • A98 CC10 • 898 CC19 • 1 CC09 • 1 CC08 • 19745 Cca8 • 27982a CC07 • 1 CC07 • ·.............................................................................. V V ••••••••••••••••••••••••••••• QUTPUT FILE DATA ••••••••••••••• PRINT DATA •••• • GROUP LINES • __•_ _ _ _ _ _ _ _ ~~~~I:iABLB];COB..lLE1ELDS/RCCL...:....RMJ.~!'PL..ffL.N_O._ _· _ • VOl.UME AVG MAX A~PH NUM AND MODE SIZE PAGE HDRS SPG: • ccn MAST MT 279.820 .. CCk'A wORK MT 39,'+90 135 135 21 • CC09 OUTP PR 0.210 __ ._TYf'f:_.....QUTP TY. ____ .. 1,0.0 • CCll 'ORK MT 1.750 _---.!__... _------ - - 132 1 135 132. 1 27 1 27 125 22d 20 TO C202 • .. bo 2 • • 25 ................................................................................ ~135 automatic output of these preliminary simulations are the systems specifications (which have been used by computer manufacturers as an ideal form for communicating processing requirements for bidding purposes). . 7. Programming Management One of the outputs produced by SCERT provides, possibly for the first time, a reasonable vehicle for the management of programming workloads. As a byproduct of its presimulation algorithms, SCERT projects - SCERT ANA~,SIS PAI'l 4 PROGRA'IMING REQUIREMEr,TS REPORT CO",PuT£R CO',Pl.EM£NT NU'.B£R Roo .. _ - - - - .. PHOGHAM _---------------- eRO.jECTIQN OF N!!MBE8 OF INSTRUCTIONS IN THE PROGRAM HUN PROGRAMMED FuRNISHED BY THE MANUFACTUHER _ _NUME;Ul. __ ....__ ~ _ _ --.5i.1BROuTlNES UTILITY ROUTINES PROGRAMMING EFFORT IN MAN MONTHS _..-J:a.tI2 ___-.---2..r.l'15_ _ _ _ ~.~74L......--------____9~4-~ C2~3 2.207 1.570 (204. 1 'e;)2 1.570 C2~5 2.192 1.718 1.154 .. '''-3.L_.________ ....272 n __ 11.7 A 7 11.5 0 9 8.000 C432 l.h4 ___ .. _ _ _ JJ..l..2l.. Dl~1 459 1.579 1.9a6 157 2.5 0152 2. ]69 1.67~ 10 5 0153 1.3b3 1.6H 6.5 .{~~~. .----t~~;~;------~~~:~5~';';;~'-----------2;;-';i'-'-.~~-- .. O,}~,,_ __ ._...1..IJo9'L- _ _ _ _..... 1.'-""u:74"--_ _ _~_ _ _ _ _"'5.... 5~_ _ 1.b84 11466 6.8 03.7 O.3k1H D321d 1,778 _Fld4 ..______ .5l"-----_ _ _ Fll" 2.192 1r466 ----"-lu"-'-_ _ _ _ _ _ _ _ _ --m~- 7.6 _ __ ~2"'2 ----t·~i~'--------'iJ..:~"-!7"'~-----------'2~~. . 6'----. 11570 10.2 1222 P10~ 2.9b9 1.400 14.8 PJ~" 2.9b9 11400 14.8 . _ _ P21L ___ -.5.t.5~.Q._ _ _ _ ___'I"".4'_,6!.Q.6_ _ _ _ _ _ _ _ ___'2:.L7.....2"___ _ _ _ ~~t. ____ ----'l.8L7_ _ _ _ _....',,'-"".1.'74'---_ _ _ _ _ _ _ _ _4......,"_ _ ._-_.---_._------------------- TOIAL. - '+b,758 223.8 the programming effort required to implement each computer run simulated. This projection can provide a standard measurement to allocate programmer workloads and to derive a degree of relative efficiency and effectiveness for a programming staff. 27 c&a WORLD REPORT - GREAT BRITAIN I saw a bored Alexei Kosygin replying perfunctorily to official speeches of welcome when he visited the Elliott-Automation development plant at Borehamwood, North of London, early in February. But ten minutes later I saw an animated, interested textile machinery engineer, Alexei Kosygin, discussing the possibilities of a computer-controlled laser beam system to cut out textile patterns at very high speed and with the greatest economy of cloth. There seemed little doubt that the Soviet Premier has a good grasp of what automation can do for industrial processing. Whether he is quite so interested in replacing his armies of bureaucrats by business computers remains to be seen. One rather piquant detail of his visit was the praise he lavished on the design of Elliott's 920 M microminiaturised computers for fire control, navigation, and other military applications. Since these computers are based on integrated circuits developed under license with Fairchild Camera, it is doubtful whether the U.S.S.R. would be allowed to buy them, even for the "Concordski." However, in the present milder climate and provided the ultimate use could be specified, who knows? software is not running as smoothly as it should, according to Dr. L. Rotherham, Member for Research. The existing IBM 7094 II is being retained for as long as necessary, Dr. Rotherham said, adding that the 360/75 "was about twice as effective." In the meantime, the Board has ordered five ICT 1905 machines, worth about $1.4m each, one for each of the Regional Centres. Interfacing these with the 360/75 is going to be quite a problem. An attempt at compatibility between the smaller 1904 of the Stationery Office and the 360-compatible System-4 computers of the Post Office is to be made by an independent programming group. A design and research collaboration agreement between Scientific Data Systems, General Electric Company of the UK, and the CEA organization in France has broken down, and GEC Computers and Automation is being absorbed by Elliott-Automation. The latte:r will work off the seven outstanding contracts involving machines from the SDS range and is unlikely, to pay much more than the value of the contracts for the goodwill, programming know-how, and sales team. One Elliott man hinted that some contracts had already cost GEC a mint of money on the systems design side. He told me that it was most unlikely that his company would continue the connection with SDS. It was not needed since Elliott is in about fifth place worldwide as a builder of process control computers. There has been a spectacular sale of a single $2Y4-million computer installation to East Germany by International Computers and Tabulators for use in business and scientific computing by the internationally-known Carl Zeiss optical works at Jena. The configuration is one of the largest 1907 s yet specified, and has a mass of tape decks, visual display consoles, magnetic card files, and plotters. With it goes $600,000 worth of data preparation equipment to the main import/ export agency in Berlin. This is the largest data processing complex ever exported from Britain, and it brings the total value of sales to the Iron Curtain countries in the past twelve months to around the $15m mark. It seems slightly ironical that, while the British computer industry is rejoicing at these export successes, the imbalance of imports into Britain against exports is still cripplingly heavy. In 1966 the import figure was about $70m, overwhelmingly IBM 30's, 40's, and 50's from France and Germany. British European Airways, extremely happy with its twin 490 (UNIVAC) seat reservation system, is going to base a four-stage integrated management system on its main computer centre. Extra equipment - computers, visual displays, booking consoles - will be needed at each annual stage. By 1970, something like $30m worth of hardware will have been utilized to control passenger schedules, crew rosters, freight, etc., and to pFovide advanced simulation powers for management's forward planning. The competition for the additional hardware has been thrown open to all comers but it is already certain that the central machines, if they prove to be overburdened (as is most likely) will be replaced by UNIVAC 494's. At the same time, IBM has scored a resounding success on the university front with the sale of a 360/67 to the joint computer departments of Newcastle and Durham Universities. Asked why he did not go for British machines, Dr. Page of Newcastle said he could not wait until the end of 1968 for an English Electric System 4-75, and he would not take an ICT machine because it had no paging facility. Dr. Page has been given specific promises of multi-access software from IBM by the end of October this year, though he did claim already that it would not be as efficient as had been hoped. The same theme comes from the Central Electricity Gen('rating Board which has taken delivery of a 360/75. The Ted Schoeters Stanmore Middlesex England 28 COMPUTERS and AUTOMATION for April, 1967 " RECOVERY FROM ERROR Jan B. Hext Basser Computing Department University of Sydney Sydney, Australia (I "A recent computer installation for handling air-traffic required that no interruption last more than 30 seconds during the first five years of operation. Such demands for nonstop service are becoming increasingly common in the range of on-line applications currently being developed." Ten years ago, if computing went wrong, recovery was a comparatively simple business. For instance, if the machine broke down, Or there was a flash of lightning, or you pressed the wrong button, the usual procedure was to start all over again. Hardware Failure Of course, if the program had been running for several hours, this was annoying. With such programs it was standard practice to dump the contents of all registers on to magnetic tape every 15 minutes or so, and to recover from breakdowns by a return to the last dump. In some installations, if there was a power cut during a magnetic tape transfer, the tape was liable to snap. If this, or any other fault, occurred while the dump was in progress, the dump itself became useless. It was therefore important to ensure that the previous dump was not overwritten until the new one had been successfully made. Ideally the dump would alternate between two tapes. The same principle still underlies all file updating techniques, in which the original file is never overwritten until the updated version has been safely established. Program Failure But hardware failures were not the only source of trouble. Far more errors were caused by program failures, and so recovery procedures had to be available for them also. On the "open shop" system, the techniques were few and simple. On failure, a red light would flash up, indicating trouble; maybe a few diagnostic tit-bits would be punched on paper tape; and then the machine would come to an abrupt halt. To cope with these regrettable emergencies there would be a set of post-mortem routines available, which the user could supply to the machine in order to retrieve further diagnostics. Then there were those programs that produced no failure at all, but quietly went into an infinite loop. Once again, with Joe standing by, this caused no great trouble. He would guess at what was happening and take on-the-spot corrective action. This might involve throwing a switch to alter the course of the program. Or maybe he could use a built-in facility, which would enable him to determine the cycle which was being repeated. COMPUTERS and AUTOMATION for April, 1967 In the age of leisure, Joe could even inspect various registers by means of lights or display tubes; he could alter little bits here and there, and then restart at any point he chose. As a last resort, he could step through his program one instruction at a time, watching to see what happened. It was all a matter of patience, not hurrying too much, being careful to press the right buttons. Naturally Joe enjoyed having full control of the works: it was a good and thoroughly desirable state of affairs. Unfortunately, however, in the middle of his debugging session, George would come breathing down his neck and he would get a little flustered. Soon he would press the wrong button, and then he would have to recover from his recovery. Batch Processing Allowing Joe to dither over which button to press, or which post-mortem to use, soon became too wasteful of valuable machine-time. Moreover, waiting while Joe made his on-thespot corrections was wasteful of George's time too. Shortage of machine-time, man-hours and patience eventually led to the batch-processing system. Under this scheme, programs were supplied to the computer in batches and then processed concurrently under control of a Supervisor routine with a minimum of outside intervention. As a result, Joe was kept away from the computer and all his programs were handled by one or two trained operators. Batch-processing had a profound effect on systems design, and not least on recovery procedures. As before, the need for recovery arose from two causes - hardware failure and program failure - and means had to be devised for coping with each. If there was a power failure, or some other such breakdown, recovery depended on knowing what program was in progress at the time and being able to restart the run at that Dr. Jan B. Hext graduated in Mathematics from Cambridge University, England, in 1962, and stayed on to take a Ph.D. in Computer Software. He recently joined the lecturing staff of the Basser Computing Department at the University of Sydney, Australia, where a network of six computers is currently being installed. , 29 I program. This was quite straightforward, as the Supervisor would normally type out the name of each program as it was entered. In some installations, to avoid back-tracking after a breakdown, means were devised for dumping all active registers immediately when a power failure was detected. A flywheel, or some other device, ensured sufficient current for the dump. The problem was then reduced to that of re-starting from a standard dump and no computation was wasted. The commonest hardware failures occurred in peripheral equipment, especially with magnetic tapes. If reading from magnetic tape produced parity failure, some units automatically made a second attempt; if the source of trouble was a speck of dust, the repeat would very probably be successful. But if a peripheral failure persisted, the hardware recorded it in a special register. It was then the programmer's responsibility to inspect this register and to take action if failure had occurred. If the Supervisor found such trouble, it would inform the operator by means of the on-line typewriter. IIKeep the System Goingll In batch-processing, the second type of recovery - recovery from program failure - had one major goal in view: to keep the system going. A program failure must not be allowed to bring everything to a halt. For example, overflow, which in early machines caused an abrupt full stop, would now be recorded in a special register which the program was left to inspect for itself. More drastic errors, such as an attempt to read cards from the line printer, returned control to the Supervisor, which would terminate the offending program and carryon to the next. It would probably print out a standard set of diagnostics by way of a post-mortem. One of the greatest dangers was that a program might overwrite pa:rt of the Supervisor. This was ,so catastrophic, and programmers attempted it so frequently, that special hardware was introduced to prevent it. By means of this hardware, the Supervisor restricted a program to an allocated area, and any attempt to stray outside the area was immediately cut short. If a program went into an infinite loop, intervention was necessary. In some systems, the programmer had to specify a time-limit to the operator; then, if this was exceeded, the operator would interrupt the system and cause the Supervisor to take over. In other cases, the programmer had to specify a time-limit in his program heading. The Supervisor would then check his run-time automatically and intervene if this was exceeded. It was important to distinguish between his "run-time" and his "elapsed time," since his program might be held up for long, unpredictable periods, e.g. in waiting for a magnetic tape to be mounted. Such a delay would not be included in his run-time. His prografll heading could also specify a limit on his output volume. This was important because it was usually possible to output vast volumes of garbage in comparatively short run-time. His program heading could also include are-start address, to which control would be transferred after certain types of failure. This allowed him to use his own postmortem routines. Third Generation Requirements The change from open-shop to batch-processing thus introduced a new level of complexity in recovery techniques. The most important aspect was that as far as possible recovery must be handled automatically by the computer itself, using a combination of hardware interrupts and Supervisor routines. Today, with "third generation" computers being installed third generation recovery problems are also being introduced. 30 The big issue now is how to recover from failures in an on-line, multi-programming environment. The problem is highlighted in situations demanding nonstop service to the users. For example, a recent installation for handling air-traffic required that no interruption last more than 30 seconds during the first five years of operation. Such requirements are becoming increasingly common in the range of on-line applications currently being developed. For them, the prospect of having to recover from breakdown is not attractive. Precautions Against Breakdown The problem becomes so acute that elaborate precautions are taken to avoid breakdown altogether. For example, one airline company with a reservations and message handling system in Sydney plans to use three independent power supplies. And should all three fail together, battery supplies will be available for an additional half hour. Similarly, if the computer itself fails in any way, or if it simply requires routine maintenance, there must be one or more reserve computers ready to take over. Of course, such back-up computers do not sit idle until a breakdown occurs: they are normally occupied with less critical work, which can be suspended when necessary. In cases where it is vital that all errors be detected immediately, two computers can be run "back to back," each duplicating the work of the other. Then, if their outputs differ, the error is immediately spotted. In an extreme case, such as rocket control, where the recovery from such errors must also be immediate, three computers can be run together and a "majority vote" taken when two of them disagree. In installations serving a number of consoles, precautions are usually more modest. The general aim is to combine the best features of open-shop and batch-processing; but great care is needed if the user's needs are to be satisfied. Time-Sharing From the user's point of view, the wheel has turned full circle. Once again Joe can bring his program to be run and retain full control of it himself. No longer need he hand it to an operator and wait for the results. Instead of the control panel, he now sits at his console and supposedly has the happy impression that the machine is entirely at his disposal. If his program goes wrong, he can correct it on the spot; if it loops, he can interrupt it or change its course. Admittedly, George is breathing down his neck once more, but all in all it feels good to have returned to open shop facilities. However, although Joe thinks he has full control of his program, in fact he does not~leven know where it is. It may be in core, or on disc, or on tape - who can tell? Moreover, he has no original program on cards or paper tape: it was all type straight into the machine, and later overlaid with corrections. So in this respect his program is more remote from him than ever. He is utterly dependent on the computer to effect all recovery procedures for him. Recovery by Computer In order to do this, the system's basic need is to know the most recent copies of all his files. These may be on disc (or drum); but if the disc itself runs into trouble, it is preferable to have copies also on tape. In one such system, the entire contents of disc were dumped on tape each day. The process took two hours, but was the only means of recovery from loss of information on disc. In other systems,.. such dumping is a continual process. Copies of all files are held on tape, together with the time and date when they were written there. Active files are also held on disc, where they can be amended and processed. COMPUTERS and AUTOMATION for April, 1967 'l Dumping is then applied as soon as possible to updated or newly created files. But even with these precautions, recovery of a large number of active programs is by no means simple. Another hazard of such systems is that in the frequent transfer of files from one device to another some information may be corrupted. The dangers and attendant recovery problems are intensified when several computers are interconnected - a situation that is likely to become quite common in the future. Although hardware incorporates many checks of its own, it may well prove necessary to include others in the software, possibly in the form of old-fashioned sum-checks. Systems Testing With sophisticated systems, a minor fault can often cause a major headache. To track it down may require a detailed investigation of both the hardware and the software. The testing of hardware is normally carried out by the Supervisor, whenever there is time available, and it is hoped that this will show up any faults; but a transient error may slip through. "Software is less amenable to such systematic checking and will usually come under suspicion first, especially as hardware is becoming increasing reliable. But since software should be checked out once and for all when written, very little attention is paid to providing facilities for locating and correcting errors in a working system: in theory they never arise. Yet even assuming that systems really are error-free after check-out, few ever remain static. New facilities will be added and existing ones modified. It may seem preferable to suspend all operations during software development; yet it is only under full working conditions that some parts will ever be fully tested. Besides, in some cases it is impossible to set aside time (~r development; systems testing and normal computation must be carried out in parallel. One way to do this is to simulate the system by means of a program within the system; then software testing can be included as a normal job. But the complications are often too great to be worthwhile. Besides, there is a danger of confusion between real failures, simulated failures and simulator bugs. Another technique is to write the system as a series of "bricks," marking some as "safe" and oth.ers as "under test," rather like a set of drugs. Those under test are automatically watched to ensure that they are not upsetting the system. If a brick becomes suspect it is replaced by a previous version which is known to be safe. This is no easy task; but something of the sort is needed if systems are to cope with the pressures of modern life. The Prospects Such are some of the problems with the up-and-coming generation of time-sharing systems. Recovery from hardware breakdown would be a gloomy prosp"ect indeed, were it not that the hardware is becoming more and more reliable. As the need for such recovery becomes correspondingly remote, and as recovery itself becomes increasingly difficult, the time may be near when the procedures will return to their earlier simplicity and attention will be focussed almost exclusively on eliminating the need to use them. But software failures seem destined to survive longer. Only recently have the difficulties of producing reliable software been fully appreciated, and for the time being we continue to improvise from one crisis to the next. The more distant prospect suggests techniques akin to those mentioned above, in which software bricks can be written and tested more systematically. But it would be preferable to ensure their reliability in the first place. As always, prevention is better than cure. COMPUTERS and AUTOMATION for April, 1967 Where would you put a record that mustn't be lost, and may be needed at any time? O.K. Where would you put 250,000 of them? And while you're at it, remember that no one has time to wait for any of them. That's the kind of problem an Acme Visible systems analyst might run into any day of the week. And every time his answer would be different. Because each company he visits has a different kind of information, a different way of using it, and a different paperflow need. Sometimes he speeds up access to automation. Sometimes he makes customer records instantly available. Another time he might have to conquer inventory control. And he does it, not by selling equipment, but by devising a unique system each time. He'll work toward saving you time, or money, or space. Or all three. Toward giving management better control, or providing faster customer service. Or both. If any of these appeal to you, call your Acme Visible systems analyst and give him the problem. He won't give you an answer. He'll design one. And it'll work faster. t&ff4IVISIBLE I Business and paperflow systems designers Offices in 76 cities in the U.S. and Canada f'l j, pfo' 8704 West AI/view Drive, Crozet, Va. 22932 '0 Co"d" Aom. S",., B,,',m S"tom,. f ~~,~"o, Designate No. 7 on Reader Service Card . 31 c&a CAPITAL REPORT The Office of the Secretary of Defense (OSD) has called for bids on a study to specify an information system to meet their expanding needs. The study is the initial step in a longrange project to develop an optimum management information system for OSD's seven Assistant Secretaries and their staffs. OSD eliminated hardware suppliers from consideration because of a possible conflict of interest, and will have the Mitre Corporation monitor the work of the selected contractor. The Secretary's office indicated that they are looking for fresh and unusual approaches to management information reporting. They invite radically different systems and concepts, and want to exploit new techniques developed but not yet operational. OSD wants the study to produce recommendations for a prototype system of radical design which could be implemented in approximately 24 months. The door was left wide open for bidders to submit their ideas, and no specific funding limit was given, but a tight December 31 schedule was set for completion of the initial study. Congress is taking an ever-increasing interest in the EDP industry as the session gets rolling in the capital. Rep. Cornelius Gallagher (D-N.J.) wants more information from the Bureau of the Budget and the Executive Branch before scheduling hearings by his House Special Subcommittee on the Invasion of Privacy. He has been openly skeptical of proposals for the National Data Center, and has voiced demands for justification from its proponents. Senator Edward Long (D-Mo.) has scheduled Senate hearings on the same subject by his Senate Subcommittee on Administrative Practice and Procedure. Rep. Jack Brooks (DTexas) is planning on holding hearings sometime in this session to look into putting the Federal budget on tape, disk, or Nth generation media, and having all agencies submit their budgets in machine digestible form. Rep. Emanuel Cellar (D-N.Y.) will have his House Antitrust Subcommittee look into antitrust and monopoly problems in the computer industry sometime during the session, but says that it has nothing to do with the Justice Department's investigation of IBM. Copywriting of computerized information routinely handled by abstracting services, libraries, and publishing companies distributing copyrighted works will be analyzed by the Copyright office. Rep. Kastenmeier (D-Wis.) asked them to get together with EDP industry leaders and find a solution to who pays what to whom and when for copyrighted data disgorged from high-speed printers. The Federal Communications Commission wants EDP industry leaders and their legal staffs, EDP users, and others to get their arguments into its hands by October 2, so that it can go about deciding if regulation is necessary in the burgeoning 32 field. Its recommendations carry significant weight in Congress, and it is concerned with questions of competition, rates, and the future adequacy of services. The House Census and Statistics Subcommittee, chaired by Rep. W. J. Green (D-Pa.) says that it is going to see what it can do EDP-wise for the Bureau of the Census as well as the National Archives and Records Service, which is a part of the General Services Administration. The Washington, D.C. Association for Computing Machinery chapter, which has many influential members from both government and industry, is taking a stand on the privacy questions surrounding the proposed National Data Center. It formed a committee and drafted a resolution for its Executive Council to look at which says that it doesn't want "promises and good intentions to be substituted for technical safeguards and effective laws: bureaucracy has a short memory." All in all, the industry is being besieged from every possible angle in Congress' rush to solve the problems largely ignored up to this time. As the spate of hearings, investigations, and controversies rage, the industry has been prompted to begin looking for some way to answer the tough problems coming at it from all sides. AFIPS invited Rep. Gallagher to a policy meeting in Warrenton, Virginia near Washington to hear his views on how it should approach some of the problems. There may be good cause for concern. Not too long ago, the New York Stock Exchange's computerized system fell down in heavy trading and an obscure bug in the program controlling final prices came to light. Scores of erroneous closing stock prices went out to hundreds of newspapers in the country, and it took two weeks to get their father / grandfather records straight. Anyone who has ever hung up the wrong master tape can visualize what could happen . . . a grain dealer in Omaha reporting the biggest sale ever to the Data Center over a remote, overloads the system. His confirmation message for data received begins, "John Doe, 12345-6789, married, scar left knee . . . . " Computer systems which match unemployed Washington, D.C. workers with jobs open in the area are being successfully used by the United Planning Organization, a non-profit agency funded by the Office of Economic Opportunity. Both the unemployed and the underemployed are being encouraged to register with UPO at ten neighborhood development centers in the Washington area. Their names, job skills, and other employment data is recorded on input forms and then put into a data bank. Similarly, jobs found by developers in the agency are recorded and entered into a separate data bank. (Please turn to page 53) COMPUTERS and AUTOMATION for April, 1967 t <, Talk's cheap at Applied Logic. Over twenty users can talk at once to Applied Logic Corporation's PDP-6 computer with its Bryant Memory System at their time-shared computation center in Princeton, N.J. They can listen, too. All beca use of the low cost and easy access of time-sharing. Scientists, mathematicians and businessmen call in from teletype units night and day, 144 hours a week. Their manmachine dialogue may take only a few seconds. But it saves them hundreds of man-hours of work. And to help the PDP-6 do its work, Applied Logic chose a 6-million character Bryant Auto-Lift Drum. They say it expands the core memory economically and gives ideal swapping and storage for the system. And interfaces with no trouble at all. They must like it! Now they're expanding with a 60-million character Bryant Model 2A Disc File. Write Bryant Computer Products, 850 Ladd Rd., Walled Lake, Mich. 48008. Ask for On-Line Application Note #3. We'll make a Bryant believer out of you, too. BRYANT COMPUTER PRODUCTS ""xL..a ~-~ COMPUTERS and AUTOMATION for April, 1967 Designate No. 16 on Reader Service Card . 33 LANGUAGE ENGINEERING Hughes Aircraft Company Los Angeles, Calif. "Is it possible to make a machine which can reproduce all the speech sounds? At present levels of development, such a machine would occupy a considerable volume and would be a programmer's nightmare. But within the next 20 years, completely automated systems will be used, acting upon oral commands and responding with verbal answers." Among the many research projects that exist today, each more fantastic than the last, few have more appeal to the imagination than the development of a machine or device which can truly listen and talk in a human language. Robots in science-fiction stories have had these functions, but our duplication of them in real life has lagged somewhat. Yet such a device would find many uses in our daily lives, and would be as much a servant to man as any of our other pieces of automatic, timesaving equipment. The Human Speaking Mechanism All peoples speak. Every known society has a spoken language. Language is a body of spoken words combined in a rather fixed fashion and understood by a considerable community. Moreover, speech involves psycho-physiological functions, the ability to make distinctive sounds that have meaning, accompanied by listening, which includes the ability to detect and recognize those sounds and to understand their meaning. The human speaking mechanism, by combining manipulations of the vocal cords, oral and nasal cavities, tongue, mouth, teeth, and lips in various ways, can make hundreds of different sounds. However, only about 40 distinctive sounds are used in speaking English. These sounds, called "phonemes," are cqmmon to a number of other languages as well.1 There are both voiced and unvoiced sounds. The voiced include the vowels and many consonants, and are produced by vibrating the vocal cords with air from the lungs. Voiced sounds contain chiefly harmonics of the frequency at which the larynx vibrates, which ranges from about 70 to 250 cycles per second (cps) for men and as high as 350 cps for women, with an average of 125 cps for men and about 250 cps for women. The unvoiced sounds are consonants formed by the breath passing lips, teeth, or tongue, and the combination of these. As a speaker talks, the vibrations set up are transmitted through the air, affecting the ear of the speaker as well as the ear of the listener. This feedback to the speaker is important because it aids in adjusting the pitch and amp liThis article is based on a report in Vectors, Vol. 8, No.3, and is used with permission from Hughes Aircraft Company. 34 tude of the speech sounds and also is believed to play a part in the thinking process. If the feedback is interfered with (perhaps through acoustical delays or loud competitive sounds) the speech will be disrupted. Reproducing Speech Sounds Is it possible to make a machine which can repr9duce all the speech sounds? Although man retained his great interest in both the phenomenon of speaking and in machines ever since the ancient Greeks, it was not until the 18th century that these two interests really began to merge. In the 17th century, Bishop John Wilkins in England published phonetic symbols portraying the vocal tract positions for alphabetic characters. In 1779 the Imperial Academy of St. Petersburg offered a prize for a machine that could make the vowel sounds. The prize was won by Christian Kratzenstein who made five tubes approximately the size and shape of the vocal passages when producing these sounds, and energized by reeds. First Speaking Machine The first successful speaking machine was built by Wolfgang von Kempelen in 1791. Working on it over a 20-year period, von Kempelen designed and constructed a mechanism capable of reproducing short phrases, as well as individual sounds. Speech was formed by manipulating me~chanical elements which simulated the essential parts of the human vocal system. The machine was restricted to short phrases because of the limitation of the windbox to produce adequate air flow. The interest in devices for simulating human sounds also continued in earnest through the 19th century. Helmholtz excited acoustic resonators by tuning forks; Willis produced vowel-like sounds by reed organ pipes; and Alexander Graham Bell went on to develop the telephone, a device that could actually reproduce speech sounds. But it was not until 1922 that an electical device was made that could produce vowel sounds, and it was not until the 1930's that the continuous synthesis of speech was accomplished by Homer Dudley. His device, the VaDER (Voice Operation Demonstrator), was exhibited at the New York COMPUTERS and AUTOMATION for April, 1967 ·~ TO LUNGS (POWER) (the Voder). Dudley built such a device and called it the Vocoder ( Voice Coder). It was the result of ten years of development. The basic principle of the Vocoder is that it detects human speech, then transmits a description of the speech signal in code, rather than the signal itself, to a duplicate machine which reconstructs the signal at the receiver. What was said comes out in an artificial but understandable and recognizable voice; only another vocoder can do this, so the system is inherently private. One of the most successful vocoders ever developed was the Hughes HC-137, currently used by the U.S. Army. Hughes has designed and constructed several models of vocoders for military and foreign governments. It is an important communication tool for military police, defense and diplomatic work and other applications requiring transfer of private intelligence. As mentioned, one of the major technical problems yet unsolved in the development of a "listening" machine is speech analysis. There are two phases or steps in this problem. The first step is to recognize the acoustic energy as speech, and not as just noise. The second step is to decipher the linguistic meaning contained in the speech signals. Analogy between human and mechanical voice production PHONEMES VOWELS Ii/as in "beet" II/as in "bit" lei as in "bait" IE/as in "bet" I rei as in "bat" la/ as in·"hot" I:) I as in "bought" /o/as in "boat" IV/as in "bull" lui as in. "boot" I A/as in "but" Irl asin "raw" J World's Fair in 1939. The Voder used two electrically-generated complex vibrations to create the sounds, a buzz representing the larynx tone (voiced sounds) and a hiss consisting )f random noise (unvoiced sounds). Either of these sounds N'as applied to a group of bandpass filters which covered the range of speech frequencies. Each filter was separately controlled, and the outputs were combined, amplified and applied to a loudspeaker to form synthetic speecJ1 sounds. /wl as in "wet" Three Functions of Speech Communication Devices Ip/as iIi "pin" It/.8.sin "tin" /kj as in u kin" Any device that is constructed with the intent of duplicating human speech communication must contend with three basic functions: (1) speaking, (2) listening, and (3) thinking (analyzing content and responding appropriately). Dudley's Voder could "make words." A microphone can "detect," but can't "listen." Modern computers can approximate some thinking processes to a certain extent, and newer mechanisms are being developed that store information and make decisions based on this information. But to date no single machine has accomplished all functions. But progress is being made and such a machine may not be far in the future. Detecting and Transmitting Speech A machine that could accurately synthesize speech was a natural descendant of the machine that could analyze speech COMPUTERS and AUTOMATION for April, 1967 Ijl as in "yet" NASALS Iml as in "ram" In/ as in "ran" l'lJI as in "rang" Ibl as in "bin" Idl as in "din" . Ig/ asin "gun" FRICATIVES IfI as in "fin" IBI asin "thin" ./s/as in "sin" 1/las in "shill" /hl as "him" in /vI as in "vat" ISI as in "thai" Iz/as in "zoo" 131 as in "measure" AFFRICATES /t las in '~chin" Id 31 as in "gin" 35 Limitations of Word Recognizers The Ultimate Speech Recognizer Several attempts have been made to construct an automatic speech recognizer. One of the first was Audrey, acronym for Bell Laboratories' Automatic Digit Recognizer. Audrey was constructed to recognize the words denoting the digits from zero to nine. Its computer has stored in it the energy-frequency patterns for these ten words, and when the machine "heard" a word, the computer would compare it with the stored patterns to obtain a match. If the patterns matched, the computer would then light a lamp corresponding to that It is likely that the ultimate speech recognizer will operate on both phonemes and words, though a problem arises in the fact that two persons saying the same word do not produce identical acoustical signals. As an approach to a solution to this problem, Hughes has continued with the design of a speech analyzer, AID (Acoustic Input Device), which could be used as the input to a linguistic processor. Much work has. been done on vocoders, speech analyzers and synthesizers, and language translators for various immediate applications. But to date no one has assembled a complete, integrated, automatic listening and speaking machine. Such a machine could be a self-organizing system, combining a speech recognizer, automatic translator, information processor, speech synthesizer, information retrieval computer, various other input-output devices, and governed by a central linguistic processor, programmed to control each of the subsystems. At present, this conglomeration would occupy a considerable volume and would be a programmer's nightmare, even if all the equipment were perfected and available as off-the-shelf items. However, as some of these devices become needed, the advance in packaging and programming will culminate in a more compact and workable system. It seems likely most of the sensing equipment will be developed to the point where it can be contained within one unit. A language machine can then be installed into any existing computer facility, using the computer as the central processor. Applications of Voice Commands Language elements 36 There are many man-machine functions which could benefit from voice-controlled machine operation, and in which voice communication would increase operating efficiency. Typewriter and punched card print-outs are fairly effective for displaying the output of most modern computers, but there are many occasions when an audible communication would allow faster machine response. In fact, any realtime information retrieval operation can be speeded up by voice control. For example, imagine using the telephone to ask a computer for a market forecast, for the answer to an engineering problem, or for the name of a book describing some obscure function, and then having the computer give the answer audibly. Another example of an activity that would benefit is air traffic control. Voice communication with aircraft by the tower operator is necessary when many aircraft are attempting to use the airport facilities almost simultaneously, and when the tower controllers are kept extremely busy tracking aircraft and talking to the pilots. A language machine, working with the data processors most large airports already have, could ease this burden. In the next few years simple versions of language equipment surely will appear as hardware and will be used for various purposes. Ultimately, completely automated systems will be used,acting upon oral commands and responding with verbal answers. At the current rate of development, such a system will probably be developed within the next 20 years. digit. But Audrey, like most other word recognizers, is limited in its ability; it is most acute to the sounds of one individual and less responsive to others. The success achieved thus far by word recognizers is limited because of their very small vocabularies. The construction of phoneme recognizers has also been attempted but with limited success. A phoneme recognizer has the advantage of dealing with a smaller number of distinct units than does a word recognizer, although the error probability is somewhat high at this level. COMPUTERS and AUTOMATION for April, 1967 CALENDAR OF COMING EVENTS " April 4-7, 1967: Honeywell H800 Users Association (HUG) Spring Conference, Bellevue Stratford Hotel, Philadelphia, Pa.; contact R. E. Hanington, Philadelphia Electric Co., 2301 Market St., Philadelphia, Pa. 19103 April 4-7, 1967: Joint Conference of the Univac Users Association and the Univac Scientific Exchange, Fontainebleu Hotel, Miami, Fla. Contacts: UUA Murray Hepple, Harris Trust, 111 W. Monroe St., Chicago, Illinois; or USE. - S. C. Bloom, Univac, P.O. Box 8100, Philadelphia, Pa. 19101 April 6-7, 1967: Atlantic Systems Conference, Americana Hotel, New York, N.Y.; contact Dr. Gibbs Myers, The General Precision Co., Wayne, N.J. April 7, 1967: Association for Computing Machinery, San Francisco Bay Area Chapter, Jack Tar Hotel, San Francisco, Calif.; contact A. E. Corduan, Lockheed Missile & Space Co., P.O. Box 504, Sunnyvale, Calif. 94088 April 7-8, 1967: The Purdue University Chapter of the Association for Computing Machinery, Purdue University, Lafayette, Ind. 47907; contact Vance H. Sutton, Chainnan, INCUM, Computer Science Center, Purdue University, Lafayette, Ind. 47907 April 12-14, 1967: Electronic Information Handling Conference, Flying Carpet Motel, Pittsburgh, Pa.; contact. Allen Kent or Orrin E. Taulbee, Co-Chairmen, Univ. of Pittsburgh, Pittsburgh, Pa. 15213 April 18-19, 1967: ECHO (Electronic Computing Hospital Oriented) Annual Meeting, American Hospital Association Headquarters, 840 N. Lake Shore, Chicago, Ill.; contact Howard Abrahamson, Director of Data Processing, Fairview Hospitals, 2312 South Sixth St., Minneapolis, Minn. 55409 April 18-20, 1967: Spring Joint Computer Conference, Chalfonte-Haddon Hall, Atlantic City, N.J.; contact AFIPS Hdqs., 211 East 43 St., New York, N.Y. 10017 April 19, 1967: Eighth Annual Southwest Systems Conference, Systems and Procedures Association, The Towne House, Phoenix, Ariz.; contact Robert V.Montopoli, UNIVAC Division of Sperry Rand Corp., 3443 N. Central, Suite 410, Phoenix, Ariz. 85012 April 20-22, 1967: Oregon Association for Educational Data Systems, Spring Conference, Portland State College, Portland, Oregon; contact Phil Morgan, Room 015,' College Center, P.O. Box 751, Portland, Oregon 97207 May 3-4, 1967: Annual National Colloquium on Information Retrieval, Philadelphia, Pa.; contact R. M. Hildreth, Publicity Chainnan, Auerbach Corp., 121 N. Broad St., Philadelphia, Pa. 19107 May 3-5, 1967: Electronic Components Technical Conference, Marriott Twin Bridges Motor Hotel, Washington, D.C.; contact C. K. Morehouse, Globe Union Inc., Box 591, Milwaukee, Wisc. 53201 May 4-5, 1967: The Montreal Chapter of the Computer Society of Canada EXPO '67 Seminar, Windsor Hotel, Montreal, Quebec, Canada; contact Raymond A. Beaudoin, Programme Committee, Computer Society of Canada, P.O. Box 1772, Station B, Montreal, Quebec, Canada May 8-10, 1967: National Rural Electric Cooperative Association's Third Annual Data Processing and Automation Conference, Executive House, Chicago, Ill.; contact Tracy E. Spencer, Management Services, N.R.E.C.A., 2000 Florida Ave., N.W., Washington, D.C. 20009 May 18, 1967: Association for Computing Machinery Technical Symposium, San Fernando Valley Chapter, Century Plaza Hotel, Century City, Los Angeles, Calif.; contact B. G. Dexter, Jr. TRW Systems, One Space Park, Redondo Beach, Calif. 90278 May 18-19, 1967: 10th Midwest Symposium on Circuit Theory, Purdue University, Lafayette, Ind. May 23-26, 1967: GUIDE International, Americana Hotel, New York, N.Y.; contact Lois E. Mecham, Secretary, GUIDE International, c/o United Services Automobile Assoc., 4119 Broadway, San Antonio, Texas 78215 May 31-June 2, 1967: A Symposium on Automatic PhotoCOMPUTERS and AUTOMATION for April, 1967 interpretation, Washington Hilton Hotel, 1919 Connecticut Ave., N.W., Washington, D.C.; contact George C. Cheng, Symposium Coordinator, Pattern Recognition Society, P.O. Box 692, Silver Spring, Md. 20901 June 12-14, 1967: International Communications Conference, Leamington Hotel, Minneapolis, Minn.; contact R. J. Collins, Dept. of Electrical Engineering, Univ. of Minn., Minneapolis, Minn. 55455 June 14-17, 1967: Annual Meeting of Council of Social Science Data Archives, University of California, Los Angeles, Calif.; contact William A. Glaser, Bureau of Applied ~o cial Research, 605 West 115 St., New York, N.Y. 10025, or Ralph Bisco, Institute for Social Research, P.O. Box 1248, Ann Arbor, Mich. 48106 June 20-23, 1967: DPMA International Data Processing Conference and Business Exposition, Sheraton-Boston Hotel, Boston, Mass.; contact William J. Horne, Conference Director, United Shoe Machinery Corp., 140 Federal St., Boston, Mass. June 26-27, 1967: Computer Personnel Research Group Fifth Annual Conference, University of Maryland, College Park, Md. (near Washington, D.C.); contact Dr. Charles D. Lothridge, General Electric Co., 570 Lexington Ave., New York, N.Y. 10022 June 26-30, 1967: 8th Annual Joint Automatic Control Conference (J ACC), University of Pennsylvania, Philadelphia, Pa.; contact G. K. L. Chien, IBM Corporation, Monterey & Cottle Roads, San Jose, Calif. 95114 June 28-30, 1967: 1967 Joint Automatic Control Conference, University of Pennsylvania, Philadelphia, Pa.; contact Lewis Winner, 152 W. 42nd St., New York, N.Y. 10036 July 31-August 4, 1967: MEDAC '67 Symposium and Exhibition, San Francisco Hilton Hotel, San Francisco, Calif.; contact John J. Post, Executive Secretary, AAMI, P. O. Box 314, Harvard Square, Cambridge, Mass. 02138 August 22-25, 1967: WESCON (Western Electronic Show and Convention), Cow Palace, San Francisco, Calif.; contact Don Larson, 3600 Wilshire Blvd., Los Angeles, Calif. 90005 Aug. 28-Sept. 2, 1967: AICA (International Association for Analogue Computation) Fifth Congress, Lausanne, Switzerland; contact secretary of the Swiss Federation of Automatic Control, Wasserwerkstrasse 53, Zurich, Switzerland Aug. 29-31, 1967: 1967 ACM (Association for Computing Machinery) National Conference, Twentieth Anniversary, Sheraton Park Hotel, Washington, D.C.; contact Thomas Willette, P.O. Box 6, Annandale, Va. 22003 Sept. 6-8, 1967: First Annual IEEE Computer Conference, Edgewater Beach Hotel, Chicago, Ill.; contact Professor S. S. Yau, Dept. of Electrical Engineering, The Technological Institute, Northwestem University, Evanston, III. 60201 Sept. 11-15, 1967: 1967 International Symposium on Information Theory, Athens, Greece; contact A. V. Balakrishnan, Dept. of Engineering, U.C.L.A., Los Angeles, Calif. 90024 Sept. 25-28, 1967: International Symposium on Automation of Population Register Systems, Jerusalem, Israel; contact D. Chevion, Chainnan of Council, Infonnation Processing Association of Israel, P.O.B. 3009, Jerusalem, Israel Oct. 18-20, 1967: Eighth Annual Symposium on Switching and Automata Theory, University of Texas, Austin, Tex.; contact Prof. C. L. Coates, Room 520, Engineering Sci. Bldg., Univ. of Tex., Austin, Tex. 78712 Nov. 14-16, 1967: Fall Joint Computer Conference, Anaheim Convention Center, Anaheim, Calif.; contact AFIPS Headquarters, 211 E. 43rd St., New York, N.Y. 10017 May 21-23, 1968: Spring Joint Computer Conference, Sheraton Park/Shoreham Hotel, Washington, D.C.; contact AFIPS Headquarters, 211 E. 43rd St., New York, N.Y. 10017 Aug. 5-10, 1968: IFIP (International Federation for Infonnation Processing) Congress 68, Edinburgh, Scotland; contact John Fowlers & Partners, Ltd., Grand Buildings, Trafalgar Square, London, W.C. 2., England 37 SOME THINGS ARE BETTER DONE OFF LlNEPaper Tape to Magnetic Tape Conversion for example; MODEL 1720 • Virtually any paper tape reader will limit the speed of your computer, but now your computer can run at full speed. • The Model 1720 uses center tape unwind for simplest possible tape loading and handling. • Performs error and validity checks. • Broad programming capability handles virtually all code conversion or format con. version requirements. • Minimal space required; any table or desk will serve the purpose. • Low cost. • Easy on tape-mangled tapes are unheard of. • Minimizes computer scheduling problems. • Nationwide service. - Low-cost Data Transmissioo; In addition to our line of converters, Digi-Data Corporation manufactures card and paper tape Data-Phone® transmission units such as the Model 2120 Magnetic Tape Receiver and the 2020 Card Transmitter shown here. Designate No. 15 on Reader Service Card DIGI-DATA .• CORPORATION •••••••••••••••••••• • DIGITAL STEPPING RECORDERS • • DIGITAL DATA HANDLING EQUIPMENT 4315 Baltimore Avenue. Bladensburg, Maryland 20710 • Telephone (301) 277-9378 38 COMPUTERS and AUTOMATION for April, 1967 r· ACROSS TH,E EDITOR'S DESK Computing and Data Processing Newsletter TABLE OF CONTENTS Applications. New Contracts New Installations. Organization News 39 41 42 43 Business News Education News. Computer Related Services. New Products. Research Frontier 44 45 45 51 52 APPLICATIONS WEATHER FORECAST FOR ONE-HALF OF THE WORLD The U.S.N. Fleet Numerical Weather Facility (FNWF) supplies over 500 forecasts each day for one-half of the world, predicting meteorology and oceanography conditions up to 72 hours in advance of its occurence. Recently five Auto-Lift Drum Memory Systems were purchased from Bryant Computer Products Division of Ex-Cell-O Corporation and installed at the FNWF, Monterey, Calif. Officer in chargeCaptai n Paul M. Wolff, U.S .N. said, "The Bryant Computer Products Auto-Lift Drum Memory Systems are used to store all the operational functions for pre-processing and post processing as well as storing all the library records for the facility." In order to make the forecasts, the FNWF collects information on current;conditions from over 4000 observation stations. They must make about six billion computations on the information received before they can relay forecasts in the form of charts and maps to several hundred operating Navy units that the FNWF services. This is accomplished by observations collected from regional teletype weather circuits, terminating i'n computers at High Wycombe, England and Fuchu, Japan. This data is then immediately transmitted by high-speed communication links to a collection center at Tinker A.F.B., Okla. From there, it is relayed together with western hemisphere data at 6000 teletype words-per-minute to the FNWF. COMPUTERS and AUTOMATION for April, 1967 Oceanographic observations and meteorological data for selected naval operating areas are collected by outlying computer networks and transmitted again by high-speed communication links to the FNWF. operator to set up the part imprecisely since it calculates and corrects for workpiece skew. The most revolutionary feature is the adaptive scheme which is used to get increased accuracy from the mechan- • ical system. Here the observed data is recorded onto magnetic tape and an Auto-Lift Drum memory by one of the three separate computer systems installed at the facility. Captain Wolff said that the memory systems allow expansion of the data processing center's capability and give the over-all system more computing speed. IBM DEVELOPS COMPUTER-LINKED PRECISION MEASURING SYSTEM A computer-controlled measuring system --- which automatically and preci sely guages two- or threedimensional parts --- has been developed by IBM Corporation at Kingston, N.Y. The measuring system is used for quality control checking of its own manufactured parts and printed electronic circuitry. The computer-controlled measuring machine (shown in the picture, above right) automatically measures parts and processes measurement data. Measurements from this Precision Measuring System-analyzed and summarized in a quality report --- can be studied a few seconds after measurements are taken. The measuring system has a number of features. It allows the The system was described in a paper -- "A Computer-Controlled Measuring System" --- presented by Alan Frane, of IBM's Kingston laboratory, at the Fourth Annual Meeting and Technical Conference of the Numerical Control Society. (For more information, designate U41 on the Readers Service Card.) HOSIERY INDUSTRY HELPED BY COMPUTER The increase in high fashion hosiery styles has meant a bigger inventory control headache for the hosiery manufacturer and retailer. 39 Newsletter L. A. Brabender, vice president and general manager of the National Mills Division of U.S. Industries, Inc., Memphis (Tenn.) Rlant, reports, "A new computer has helped us cure this headache and has saved us money in the bargain." National Mills is one of the nation's largest hosiery manufacturers and markets stockings under its own Roundthe-Clock brand, as well as many leading private label brands. A computer-based store stock replacement system has been evolved which keeps customer store inventories at a maximum service level to meet shopper demand. Thesystem also regulates the distribution of hosiery from National's distribution plant at Memphis and schedules stocking production at the company's mills in Grenada and Batesville, Miss. Mr. Brabender says National's computerized stock replacement program was a first in the apparel industry. The IBM System/360 Model 30 computer provides data on what is being sold and where; it can predict hosiery sales trends by region and ev~n by individual store. To set off the merchandising control cycle, an IBM card is placed in each box of hosiery shipped by National to a store customer. When the box is opened for sale, the card is returned to National's headquarters and processed via the computer to initiate an immediate replenishment order. Computerprepared information tells packaging personnel at the firm's Grenada, Miss., mill exactly what stockings to package, which labels to put on the boxes and what packaging materials to use. At the same time, the computer keeps track of the store's shelf stock bY'style and color and of the mill and warehouse inventory supplies. Because the computer stores data on what has been sold the previous week by stores across the country, it can predict the stockings that should be knit, dyed, packaged and shipped this week and the next. The computer also considers seasonal trends and other sales variables to predict customer demand for styles and colors months into the future. ASE HAS COMPUTER-BASED DATA RETIUEVAL SLSTEM A computer-based data retrieval system now enables officials of the American Stock Exchange and member firms to obtain immediately 40 price range and volume data on stocks and warrants traded during the latest six weeks. The new procedure replaces a 35-year old manual operation. The new electronic system was developed in cooperation with The Bunker-Ramo Corporation and is called Record Room Interrogation System (RRIS). Data is made available to record room personnel on a BunkerRamo Cathode ray tube input/output device (BR 203). In addition, on push-button request, the RRIS will simultaneously prepare a print-out summary of stock information. Range and volume records for the six latest trading weeks are maintained at Bunker-Ramo's New York City data center (TeleCenter) on storage drums of a Univac 418 computer, from which it is called out instantaneously through the BR 203 in response to a telephone call to the Record Room from a member firm. Query of records is made through the BR 203 device which has seven function keys for seven separate types of market data. There are two inquiry devices in the Exchange's Record Room and another in the Stock Watch section. The Record Room Interrogation System has been designed to compile trading records from time of entry at the Trading Post. The system is programmed directly from the ticker tape; a TeleCenter computer digest tape data; posts the data on magnetic tape; and transfers the material to the Univac 418. These functions are performed at the close of each business day, so that trading data is ready for dissemination the following morning. COMPUTER ANALYZES CHEMICAL SAMPLES AT MONSANTO PLANT Some 1000 chemical samples a week now are being analyzed by computer at Monsanto Company's Chocolate Bayou Plant near Alvin, Texas. Use of the computer, and IBM 1800 data acquisition and control system, has dramatically improved the precision of laboratory analysis and reduced the time required for completion of tests. A major improvement is in the precision of analysis achieved. Average improvement of results has been tenfold. In one instance, sampling with a variation of 4 per cent in results was changed to a variation of only .02 per cent. Another saving is in the time required for analysis. Computercontrol permits a faster run through the chromatograph (an analytical instrument which determines the nature and concentration of compounds in a sample). Significant time also is saved in calculation of results. A chart with 100 chromatographic signals takes 2Yz hours to interpret manually. The computer does the job in 3Yz minutes. The system, jointly developed by Monsanto and IBM, monitors 40 gas chromatographs. It is designed for routine operation by laboratory personnel. In the picture below, a lab technician is shown using a table-top keyboard to tell the comI puter what technique to follow, The IBM 1800 computer, located adjacent to the control lab, is programmed to run the sample through a testing instrument and interpret its data. Results are printed on an output typewriter and furnished manufacturing personnel. Although several computerized laboratory systems have been developed recently, the system at Monsanto is said to differ significantly in several respects. The system carries out a completemethod of analysis, including external standards, internal standards, normalization and grouping. It is programmed to handle 150 methods and permits programs to be changed without taking the computer off its regular work. It is also selfdocumenting and capable of keeping its own files. No external records are required. The system's flexibility also allows the addition of essentially any advanced type of analytical method. Products of the plant include ethylene, propylene, benzene, naphthalene, phenol, alkrlbenzene, acrylonitrile and butadiene. COMPUTERS and AUTOMATION for April, 1967 , Newsletter G-E USING COMPUTER TO AID IN-HOUSE MANUFACTURING OPERATIONS computer even produces the N/C control tapes for the machines which will run the job. At General Electric Company's Computer Equipment Department in Phoenix, Ariz., a computerized methods engineering system for sheet metal production may portend a significant new development for metalworking shops both large and small. The Phoenix plant produces a broad range of sheet steel chasis, panels and other components for G-E electronic computers. More than 45 tons of sheet steel are processed each week, on a 6-day, 24-hour per day schedule. Production is handled by three numerically-controlled A-15 turret punch presses supplied by The Wiedemann Division of the Warner & Swasey Company. Lots range from 1 to 1000 pieces. William L. Lord, Manager of Shop Operations at Phoenix, says, "Regenerative planning with the computer provides the best production plan for a particular set of production conditions. It is much faster than the old method of individual manual job planning which sometimes required a cycle time of a week to ten days to complete. With the computer it is virtually an overnight service." To handle production control and raw material inventory for this complex metalworking operation,G.E. uses one of its own G-E 400 line computers. All of the various production parameters and workpiece specifications are stored on magnetic tape. When a job order comes to the shop for a particular component, the data tapes enable the computer to "regenerate" a production plan, including paper work, for the job under optimum conditions for the individual lot size and current shop conditions. The AIR FREIGHT CONTROL SYSTEM DEVELOPED BY AIRBORNE FREIGHT CORPORATION A new system for the control of air freight shipments, in which an advanced data processing operation is integrated with an extensive network of private communications lines, has been announced by Airborne Freight Corp., San Francisco, Calif. J. D. McPherson, president of the international air freight firm, said that the system is the first of its kind in the industry and represents a major breakthrough in the control of air freight, promising an entirely new dimension of service to air freight users. The total control system is named DART, from the initials describing its basic function, Direct Airbill Retrieval and Transmission. With DART, all Airborne stations, located throughout the nRtion, are directly connected with an {BM 360 computer, in San Francisco, via more than 43,000 miles of Airborne's private communications lines. After more than three years of developmental work by Airborne air freight traffic specialists, IBM data processing technicians, and Pacific Telephone communications engineers, DART now is operational. More than 220 programs have been written to instruct the IBM 360 computer. Essentially every step of each shipment handled by Airborne will be controlled or influenced by DART. Some of DART's advantages are: computer-accurate airbills, automatic alerts, instant proof of delivery, positive followthrough and fast tracing. Via DART, every station in the Airborne system can query the computer on the status of any shipment, during any portion of its journey, and recei ve a reply through retrieval, from the 360' s memory system, of the detailed information fed into the computer at every significant point in the move. The economics of DART indicate that no increase in rates wi 11 be required, the Airborne president said. NEW CONTRACTS AMOUNT Federal Aviation Agency Raytheon Company, Wayland Laboratory, Wayland, Mass. NADGECO Limited, London, England Cubic Corp., San Diego, Calif. U.S. Air Force Sylvania Electric Products, Inc., a subsidiary of GT&E, Needham. Mass. Sylvania Electric Products, Inc., a GT&E subsidiary Air Force Systems Command's Electronic Systems Division, Hanscom Field. Mass. Recognition Equipment Inc., Dallas, Texas Scientific Data Systems, Santa Monica, Calif. Grumman Aircraft Engineering Corp. Fairchild Space and Defense Systems, Syosset, N.Y. Trans World Airlines, Inc. Recognition Equipment Inc., Dallas, Texas COMPUTERS and AUTOMATION for April, 1967 Developing, producing, installing and testing automation equipment for modernizing the nation's long-range, air route traffic control centers; the equipment, described as the Computer Display Channel (CDC) of FAA's advanced National Airspace System (NAS), is a key element in the Agency's emerging semi-automatic air traffic control system Production of 37 computer peripheral systems for the NATO Air Defense Ground Environment system (NADGE) Expanding the data processing capabilities of Minuteman II intercontinental ballistic missile systems Installation and system testing of special purpose electronic receiving equipment Twenty SOS 910 computers to be used in Recognition Equipment's Electronic Retina® Computing Readers Design, development and preproduction quantities of data converters for the avionics system of the EA-6B aircraft An Electronic RetinaW Computing Reader to automate computer data input $44.8 million $3.7 million $3 million $2.1 million $1.6 million $1,449,914 about $750,000 41 Newsletter International Computers and Tabulators, Ltd., London, England Data Products Corp., Culver City, Calif. Data Products Corp., Culver City, Calif. Oak Park and River Forest High School, Oak Park, Ill. Houston Lighting & Power Company, Houston, Texas Ampex Corp., Redwood City, Calif. Leeds & Northrup Co., Philadelphia, Pa. United States Steel Corp. C-E-I-R, Inc., Washington, D,C, Aro, Inc., Tullahoma, Tenn. Systems Engineering Laboratories, Inc., Fort Lauderdale, FIa. Board of Education, Orange, N.J. Great American Insurance Companies, New York, N.Y. System Development Corp. (SOC), New York Metropolitan Operations Center. Paramus. N.J. Mohawk Data Sciences Corp., Herkimer, N.Y. American National Insurance Co., Galveston, Texas Ampex Corp., Redwood City, Calif. Ampex Corp., Redwood City, Calif. Model 5045 DISCfILE random access memory over $420,000 systems to be used in conj unction wi th the new advanced ICT Atlas computers 100 Model TM-7 digital tape units which $400,000 will be incorporated into'the line of LIP MIOOO and Lip M640 Off-Line Print Stations manufactured by Data Products Manufacture and installation of an audio $358,000 teaching system A computer control system encompassing, for the first time, a coordinated overall system operation including both energy dispatching and transmission dispatching through one integrated system. The system will be built in cooperation with the Philco Houston Operations of the Philco Ford Corp. Development of special network planning and mathematical programming "packages" for U,S. Steel's new 88500 computer An SEL 810A computer system to be used for total data acquisition of wind tunnel testing of the C5Aengine for the U.S. Air Force; delivery scheduled for May to Arnold Air Force Station, Tenn. Technical assistance in the planning of an Educational Resource Center 65 Mohawk Data-Recorders to be installed in Great American's home office and 16 regional offices throughout the U.S. A Videofile document filing and retrieval system to automate the handling of insurance policy files NEW INSTALLATIONS Continental Telephone Corp., St. Louis, Mo.; Bakersfield, Calif.: Syracuse, N.Y. Four of six comp~ters; three Honeywell 200's and three Honeywell 1200 systems; total value about $3 million Michigan State University, East Lansing. Mich. Potter Instrument Co., Inc., Plainview, N.Y. Sigma 7 B. B. Walker Shoe Company, Asheville. N.C. Irani Associates, Inc., Miami, Fla. Archdiocese of Brooklyn Data Center, Health & Hospitals Division of Catholic Charities for the Counties of Brooklyn and Queens Martin Marietta Corp., Denver, Colo. Tokyo ShibauraElectric Co. (Toshiba), Tokyo. Japan Mohawk Data Sciences Corp., Herkimer, N.Y. Rex Chainbelt Inc., Milwaukee, Wisc, Palo Alto Data Center, Control Data Corp., Palo Alto, Calif. First Federal Savings and Loan Association, Charleston, S.C. 42 BIT 480 general purpose computer NCR 315 computer system IBM 1130 (equipped with COGO language) RCA 301 computer Control Data 6400 computer system GE-635 computer system Honeywell 200 computer The first of six System/360 Model 20 computers Control Data 3800 computer system NCR 315 computer system Computing rates for long-distance and special calls, and for perfo~ming all customer billing, inclUding toll calls, local service charges and special service charges; later the system will be expanded to perform general accounting, data communications, forecasting of demand for construction planning and trouble analysis applications Operating the controls of MSU's 55-million electron volt nucleus smasher Use in developing special systems that will link Potter's data processing peripherals and allied products with a central processor such as the BIT 480 Order processing, billing, inventory, sales analysis. cost and general accounting Engineering calculations, construction costs analysis. etc. Providing detailed information data to doctors on patients in six New York hospitals: also hospitals' payroll and personnel records will be computerized along with routine accounting procedures A variety of technical and scientific data processing applications, including research and development. design analysis. and production support Scientific and engineering calculations, and some business data processing applications General accounting and production control operations Primary stage in the company's development of an advanced management system Increased capacity and capability; Data Center provides computer services on a rental basis to business firms and institutions Savings, mortgage, payroll and general ledger applications COMPUTERS and AUTOMATION for April, 1967 Newsletter Rock-Mill Inc., Rockford, Ill. IBM 1130 computer Bonneville-Sylvan Life Insurance Co., Salt Lake City, Utah Fort Worth Division, General Dynamics Corporation Honeywell 200 computer EAI 8800 Analog/Hybrid Computing System and an EAI 640 Digital Computing System valued at over $500,000 ORGANIZATION NEWS ENCYCLOPAEDIA BRITANNICA ACQUIRES INTEREST IN EDUCATIONAL TECHNOLOGY FIRM Encyclopaedia Bri tannica, Inc., Chicago, Ill, has acquired an interest in TECHNOMICS, Inc., a one year old California company engaged in the application of computer-centered technology to the fields of general education, governmental administration, public health, welfare, and medical programs. The agreement involved a stock purchase in which Bri tannica bought 200 shares of co~vertible preferred stock with the option to buy an additional 100 shares. All stock now i s contro lIed by TECHNOMICS and Britannica. Price was not disclosed. TECHNOMICS, Inc., was formed in December, 1965, by a group of psychologists, educators, system and computer experts from the RAND Corporation and the System Development Corporation. They have developed numerous computer applications in education and training and have applied system analysis, cost-benefit analysis, operations research techniques, gaming and simulation to the problems of development of large systems for military, health, education, and public welfare needs. President and chairman of the board is Dr. Norton F. Kristy. MEMOREX ACQUIRES DISC PACK CORP. Memorex Corp., Santa Clara, Calif., has acquired the balance of the outstanding equity of Disc Pack Corp., Laurence L. Spitters, President of Memorex has announced. Previously, Memorex had owned 40% of Disc Pack Corp. Acquisition of the balance involved an issuance of 15,000 shares of Memorex Common Stock or 1.5% of the total Memorex shares outstanding. COMPUTERS and AUTOMATION for April, 1967 Numerical control contouring work; prepares coded tapes which control milling and other precision machines Life insurance and service bureau data processing work Simulation of complex aerospace weapons systems; initially in developments for the F-lll family of aircraft Memorex is the nation's second largest manufacturer of precision magnetic tapes for computer, instrumentation and television recorders. Disc Pack Corporation, located in Hawthorne, Calif. is presently engaged in the development of preci sion memory di sc packs for use as information storage devices for IBM and other new computers. Production of disc packs is expected to begin later this year. Group, Inc., has acquired controlling interest of a French planning firm called IRCOM. IRCOM, established in 1963, specializes in economic and planning studies, market research, and construction feasibility studies. A substantial portion of the firm's business consists of planning assignments conducted on behalf of the French government. One of the current assignments of IRCOM is a large study of the impact of automation on the French economy being conducted on behalf of the French government. LEAS CO TO ACQUIRE DOCUMENTATION, INC. Diebold France, S.A. is jointly owned by The Diebold Group, Inc., New York, N.Y., and Rothschild Freres, the senior bank of the French Rothschild Group. The Diebold Group, an international management consulting company specializing in the business applications of advanced technolog~ maintains offices in six European cities. Saul Steinberg, President of Leasco Data Processing Equipment Corp., Great Neck, N.Y. CASE) and Eugene Miller, President of Documentation Incorporated ·of Bethesda, Md. (OTC) announced that the two companies have reached an agreement in principle for a pooling of interests whereby Leasco would acquire Documentation Inc. The transaction will be based upon an exchange of shares at the rate of one share of Leasco stock for each seven shares of Documentation stock. Leasco is in the business of leasing computers and other data processing equipment. The business of Documentation Inc. is the development and operation of computer information technology systems for government and industry. The present officers and staff of Documentation Inc. will continue to manage its business. The officials of the two companies stated that the completion of the transaction was subject to the execution of definitive contracts, the approval of both boards of directors and of shareholders in the case of Documentation Inc. STANDARD REGISTER SIGNS KNOW-HOW AGREEMENT WITH GERMAN FIRM The Standard Register Company, Dayton, Ohio, has concluded a knowhow agreement with Hessdruck, a printing company located in Braunschweig, Lower Saxony, Germany. Hessdruck is the third largest business forms printing house in Germany. In business forover 65 years, it produces continuous business and transportation forms and also has flat printing facilities, serving all Germany. DIEBOLD FRANCE ACQUIRES FRENCH PLANNING FIRM Under the terms of the agreement, Standard Register will provide the German firm with technical and administrative advice and assistance relative to the production of manifold business forms and the printing equipment necessary. Hessdruck wi 11 also have marketi ng right s in Germany for the sale of Standard Register mechanical forms handling equipment. Diebold France, S.A., the French subsidiary of The Diebold The signing brought the total of foreign firms having know-how 43 Newsletter agreements with Standard Register Company to 14. NCR PLANS TO ACQUIRE MICROCARD CORPORATION The National Cash Register Co., Dayton, Ohio, announced that it has entered into an agreement to acquire the Microcard Corporation of West Salem, Wisc., in exchange for an undisclosed number of shares of NCR common stock. Microcard, a privately held company, is a leading producer of microfiche information storage systems. Robert S. Oelman, NCRchairman, said acquisition of the Microcard Corporation would complement NCR's recently announced line of PCMlro microform systems (see Computers & Automation, January, 1967, p. 56). The agreement to acquire Microcard is subject to certain tax rulings and audits, the NCR chairman said. Under its terms, Microcard would become a wholly owned subsidiary of NCR and would remain under its current management headed by A. L. Baptie, president. DATANAMICS CORPORTION OF AMERICA ACQUIRES COMPUTER PERSONNEL CONSULTANTS, INC. systems engineering and programming services. DSA's activities also will include conventional EDE consulting services. Data Systems Analysts, Inc., corporate headquarters is located in Pennsauken, N.J., with regional offices located in River Edge, N.J.; Phoenix, Ariz.; Amsterdam, Holland; and Paris, France. DATA SYSTEMS ANALYSIS, INC. ESTABLISHES HEADQUARTERS IN PARIS, FRANCE Data Systems Analysts, Inc., announces the opening of its European Headquarters in Paris, France. These facilities will provide online, real-time and communication 44 An institute for advanced computer research and education at postgraduate, professional levels has been announced by the IBM World Trade Corporation. It is located in Geneva, Switzerland and begins full activity this year. About 80 of IBM's top systems engineers in Europe will be accepted by the institute in 1967. EDUCATION NEWS COMPUTERS IN THE CLASSROOM BECOMING ROUTINE IN NEW YORK CITY Students at Jamaica High School, Jamaica, N.Y., like hundreds of other boys and girls in New York City high schools, are studying computer mathematics. The New York City Board of Education reports that 22 of its high schools now offer this course to juniors and seniors. According to Dr. Bernard E. Donovan, superintendent of schools, the city's high school system makes more extensive use of computers in the classroom than any other high school system in the nation. This use is being expanded experimentally to include junior high school students. Datanamics Corporation of America (Chicago, Ill.), a data processing service organization, has acquired Computer Personnel Consultants, Inc., also of Chicago. Datanamics president, David S. Pemberton, in announcing the acquisition, said, "By adding the proven procurement ability of Computer Personnel Consultants to our present array of data processing skills and equipment, we can now offer total data processing service to our customers. " CPC is an established name in the personnel consulting and recruiting field. It has developed a specialized practice in the data processing, operations research, mathematical sciences and related personnel areas. SYSTEMS RESEARCH INSTITUTE ESTABLISHED IN EUROPE BY IBM WORLD TRADE In the picture, Stephen Orphanos, a mathematics teacher at Jamaica High School, Queens, is working with a group to program the IBM 1130 in the foreground. The student at the left, Steven Gabriel, already has written a number of programs for the computer. One of them allows the IBM 1130 to speed through many of ~he school's accounting jobs. The IBM European Systems Research Institute will extend studies of theoretical computer concepts, provide opportunities for research, and offer courses in advanced mathematics, the design and operation of computer systems, and other subjects. A permanent staff of teachers will be assisted by visiting lecturers from IBM's Systems Research Institute in New York and other professional IBM people in Europe and the United States. P-rofessors from leading universities in Europe also will lecture from time to time. ACLS AND NYU ESTABLISH A NATIONAL CENTER FOR BIBLIOGRAPHIC DATA PROCESSING The American Counci 1 of Learned Societies (ACLS), a federation of 32 scholarly organizations representing modern languages, literature, history, philosophy, religion, the arts and humanistic elements of the social sciences, is working with New York University to develop and exploit computer technology specifically for researchers in these fields. With the aid of a $144,000 grant from IBM Corporation, the ACLS and NYU have established a national Center for Bibliographic Data Processing. Dr. Frederick Burkhardt, ACLS president, stated that the work of this center will be to "tailor computer technology to the specific needs of humanists and give them the help and multiplied capabilities that computing has given researchers in' the sciences and other fields." Under the cooperative program, NYU's Institute for Computer Research in the Humanities (ICRH) , an existing research center and national clearing house for information in this field, will supply computer indexing and other special services for the publications of ACLS societies and other interested scholarly organizations. Through COMPUTERS and AUTOMATION for April, 1967 Newsletter the societies and their publications, the ACLS will encourage humanities scholars to take full advantage of computer technology in their research and study. It also will speed the distribution of developments in computing for the humanities to the academic community and to computing centers throughout the nation. copper technology, watch a data display screen as the computer was querierl, and then, within seconds, see the answer displayed on the screen in the form of reference numbers. The engineer then is able to turn to the volumes containing extracts of the documents referred to, which show the actual data in which he is interested. The major NYU project is the compilation of a massive electronic data bank of cross-referenced indexes for 30 cooperating scholarly journals. Only separate annual or cumulative indexes for individual journals are available now. This cross-referenced data bank, Dr. Burkhardt said, ultimately will provide far more comprehensive bibliographic information than scholars normally obtain in their own li brary searches. "One day", Dr. Burkhardt said, "it may be possible to keep individual 'interest profi les' on record for large numbers of scholars and periodically or on request supply them with all information in the data bank matching their specific interests." NYU researchers now are trying to overcome the technical barriers to such a system. The demonstration unit, with a desk-size computer showed -- on a small scale -- exactly how the retrieval system at CDA's Technical Data Center actually operates. The Center, established in 1965, is at the COlumbus (Ohio) Laboratories of Battelle Memorial Institute. The CDA Technical Data Center's objective is ta provide engineers, who select and apply materials, with complete and up-todate technical data on the properties, processing and applications of copper, brass and bronze. The copper and brass industry Technical Data Center, and the computerbased services it provides, are the first of their kind in the metals industry. COMPUTER RELATED SERVICES Engineers needing data of this type may make use of the facility simply by requesting specific information from CDA, CDA member companies, or the CDA Technical Data Center. By completing a CDA Interest Check List Card, engineers obtain periodic reports and data sheets in their fields of interest which are generated by the CDA Technical Center. There is no charge for either of these services. Information and data from both published and unpublished sources are collected, reviewed by speNYU investigators also are cialists at the Columbus laboratrying to develop computing "shorttories of Battelle Institute, and cuts" for humanists similar to the selected documents are evaluated special programs that handle the by one of seventy engineers in the findings of square roots and other industry who are the staff ~f exroutine problems for engineers and perts that guide the program techscientists. Dr. Jack Heller, dinically. After evaluation, the rector of the Heights Academic Cominformation and data are edited puting Facility and a prime mover into an "extract" containing all behind the establishment of the ICRH of the text, tables and figures said, "We want to give musicolojudged to have long-term usefulgists, linguists and other humaniness. This document then receives ties scholars similar computer a serial number which is stored in 'software' to handle the routine the computer memory behind each procedures in their work." appropriate term of the more than five thousand terms in CDA's "Thesaurus of Terms on Copper Technology". COMPUTER ANSWERS QUESTIONS ON COPPER & COPPER ALLOYS At the 1967 Insti tute of Electrical and Electronic Engineers Show in New York last month, a special computerized information retrieval system provided instantaneous access to worldwide technical literature on the technology of copper and its alloys at the booth of Copper Development Association, Inc. Electrical and electronic design engineers were able to ask questions on specific phases of COMPUTERS and AUTOMATION for April, 1967 NEW PRODUCTS -- D igital fRIDEN INTRODUCES 5610 COMPUTYPER DATA PROCESSOR A third-generation business machine,known as the 5610 Computyper® Data Processor, has been developed by Friden, the business machine division of The Singer Company, San Leandro, Calif. This smallscale data processor uses microintegrated circuits for speed, compactness, and reliability and has many of the capabilities of larger, higher-priced machines. The 5610 has two separate memories. For data storage, it has sixty registers (plus additional capability with auxiliary units). Totals can be read out at any time and in any format. The second mem- 45 Newsletter ory contains 1118 alphanumeric characters exclusively for internal program storage. Programming is speeded up and simplified by the 5610'8 own universal programming language, SWIFT (~oft~are Implemented Iriden !ranslator), a language composed of 38 simple alphabetic command statements, such as add, divide, branch, and type. Programs can be written in everyday English (or German, FrenCh, etc.). The 5610 itself automatically translates SWIFT English into machine language. MATHATRON 4280 An all new, desk size computer/calculator has been introduced by Mathatronics, a division of Barry Wright Corp., Waltham, Mass. This compact, low cost digital machine, designated Mathatron 4280, is designed to solve complex mathematical problems in engineering and scientific applications. Mathatron 4280, wi th over 4000 bits of storage, accepts standard algebraic and decimal numbers and includes automatic operators for square root, log, anti-log, sine, arc-tangent and parenthesis. Up to 82 individually addressable storage registers are available with internal programmable memory of 480 program steps. Master programs can automatically branch to sub programs, execute them and then continue along the original path. -- To the left of the Input-Output Station is a horizontal tape transport. The electronic processor, below the transport, is the control center for the complete 5610 system The 2205 Input/Output Station transmits data to the electronic processor and prints out finished documents. Input is automatic through punched tape, edge-punched cards, tab card, or program control. Variable data can be manually entered via its electric typewriter keyboard. Output is in the form of printed documents and byproduct tape/cards which can be used for further processing. The 5610 has its own diagnostics capability. If a malfunction should occur, a Friden product service specialist merely loads a special diagnostics program and the 5610 searches for the fault and types out the location. Corrective action can be taken quickly with minimum downtime. The new Friden data processor is geared primarily, but not limited to, business applications. Fiv~ auxiliary on-line input/output devices are available to expand the capabilities of the 5610 system to meet individual applications. These include a Tape Reader, Automatic Card Reader, Selectadata
Source Exif Data:File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.3 Linearized : No XMP Toolkit : Adobe XMP Core 4.2.1-c043 52.372728, 2009/01/18-15:56:37 Producer : Adobe Acrobat 9.1 Paper Capture Plug-in Modify Date : 2009:03:20 17:17:21-07:00 Create Date : 2009:03:20 17:17:21-07:00 Metadata Date : 2009:03:20 17:17:21-07:00 Format : application/pdf Document ID : uuid:e594376f-bcc5-4f3f-9b22-1022dcea4767 Instance ID : uuid:56d3e6dd-86d4-465c-aa5b-71cd7fd3d912 Page Layout : SinglePage Page Mode : UseNone Page Count : 60EXIF Metadata provided by EXIF.tools