|
Before the advent of electronic computers, data processing was performed using electromechanical devices called unit record equipment, electric accounting machines (EAM) or tabulating machines. A data processing shop would have at least one of the most of the machine types. Data processing consisted of feeding decks of punch cards through the various machines in a carefully choreographed progression. The flow of card decks between the machines was typically hand-drawn on large sheets of paper using standardised symbols for the various functions. Unit record machines were as ubiquitous in industry and government in the first half of the twentieth century as computers became in the second half. They allowed large volume, sophisticated, data-processing tasks to be accomplished long before modern (electronic) computers were invented. This data processing was accomplished by processing decks of punched cards through various unit record machines in a carefully choreographed progression. This progression, or flow, from machine to machine was often planned and documented with drawings that used standardised symbols for the various machine functions – drawings that today would be called flowcharts. The machines all had high-speed mechanical feeders to process from around one hundred cards per minute, to 2,000 cards per minute, sensing punched holes with either electrical or optical sensors. The operation of many machines was directed by the use of a removable control panel. Initially all machines were constructed using electromechanical counters and relays. Electronic components were introduced on some machines beginning in the late 1940s. History Herman Hollerith developed punched card and unit record technology for the 1890 census and founded the Tabulating Machine Company (1896) which was one of three companies that merged to form Computing Tabulating Recording Corporation (CTR), later renamed IBM. IBM manufactured and marketed a variety of unit record machines for creating, sorting, and tabulating punched cards, even after expanding into computers in the late 1950s. IBM developed punch card technology into a powerful tool for business data-processing and produced an extensive line of general-purpose unit record machines. By 1950, the IBM card and IBM unit record machines had become ubiquitous in industry and government. The warning often printed on cards that were to be individually handled, "Do not fold, spindle or mutilate," became a motto for the post-World War II era (even though though many people had no idea what spindle meant). The largest supplier of unit record equipment was IBM and this article largely reflects IBM practice and terminology. Punched cards The basic unit of data storage was the 80-column punched card. Each punched column represented a single digit, letter or special character. Data values consisted of a field of adjacent columns. An employee number might occupy 5 columns; hourly pay rate, 3 columns; hours actually worked in a given week, 2 columns; department number 3 columns; project charge code 6 columns and so on. Keypunching Original data was usually punched into cards by workers, often women, known as key punch operators. Their work was often checked by a second operator using a verifier machine. Cards were also produced automatically by various unit record machines and later by computer output devices. Sorting A major activity in any unit record shop was sorting decks of punch card into the proper order as determined by information punched in the card. The same deck might be sorted differently depending on the processing step. Sorters, like the IBM 80 series Card Sorters, sorted an input deck into one of 13 output bins depending on which hole was punched in a selected column. The 13th bin was for blanks and rejects. Sorting an input deck into ascending sequence on a multiple column field, such as an employee number, was done by a radix sort. Data processing tasks typically ran on a daily batch processing cycle. All the data cards punched during the day were sorted and merged with a master deck, which was then tabulated. Tabulating Reports and summary data were generated by accounting or tabulating machines. The sorted deck was fed through the tabulating machine and each card was printed on its own line. Selected fields from each card were added to the value of one of several counters. At some signal, say a card with a special punch indicating it was a master card, a summary line would be produced containing the summed values. For many applications, the volume of paper produced by tabulators required other machines, not considered to be unit record machines, to ease paper handling. Card punching Card punching machines included: Later "document origination machines" such as the IBM 519 could perform all of the above operations. Collating and interpreting A collator had two input hoppers and four or more output hoppers. These machines could merge or match card decks based on control panel's program. An interpreter would print the values of columns along the top of the card. The columns to be printed could be selected and even reordered based on the machines control panel wiring. Programming
Unit record equipment in the computer age Early computer installations used punched cards for program entry and storage. A typical corporate or university computer lab would have a room full of key punch machines for programmer use. An IBM 407 Accounting Machine might be set up to allow newly created or edited programs to be listed (printed out on fan-fold paper) for proof reading. An IBM 519 might be provided to reproduce program decks for backup. The 519 could also punch sequential numbers in columns 73-80 of COBOL or Fortran program decks. Those languages and others did not use those columns; the use of only 72 columns is a tradition tracing back to the IBM 704 card reader. An IBM 80 series sorter would be used to put things back in order if a sequenced deck was dropped. A quicker, but less effective, protection against dropped card decks was drawing a diagonal line across the top of the deck with a marking pen. Early mid-sized commercial computers, such as the IBM 1401 were designed to work with punch card operations and allowed more complex reporting. However many shops soon began using magnetic tape as their primary storage medium, using cards primarily for data input. Many organizations were loath to alter systems that were working, so production unit record installations remained in operation long after computers offered faster and more cost effective solutions. Specialized uses of punch cards, including toll collection, microfilm aperture cards, and punch card voting, keep unit record equipment in use into the twenty-first century. The IBM System/3, the original ancestor of the entire IBM midrange computer product line, was developed as a replacement for plugboard-programmable unit record machines. Modern-day applications of unit record concepts Although unit-record equipment is no longer used, the underlying concepts of unit-record-style processing are still widely used when processing very large numbers of records on tight production schedules. (Consider, for example, the processing of telephone-call records as the input to a billing system; a task which might easily involve tens of millions of records per day, or more. The task is very essential to the operation of the business, and tomorrow there will be another (say) ten million records to do...) The essential technique revolves around the use of sorting, and subsequent processing of the sorted file that takes advantage of the fact that the file has been sorted. Sorting techniques have been extensively studied in computer science and are extremely efficient. When a file is sorted, not only are the key values placed in a predictable collating sequence, but it is now known that: For example, in a file sorted by zip code, given the stream ...12345 12348 12348 12460... you can see that the key 12348 occurs exactly twice and that the key 12352 does not occur at all. Notice that you can reach these two conclusions with certainty, just by looking at these records, even if this were part of a stream containing millions of records, because the file is known to be sorted by zip-code. In fact, unit-record processing programs generally only have to consider two records ... "this" one and the "previous" one ... to make all of their decisions.) The Unix utility uniq works in this fashion.When a sorted file of transactions is used to update a master file that is sorted the same way, as in (say) a merge algorithm, both files can be processed sequentially; that is to say, random access is not needed, and expensive seek time is avoided. When millions, or even billions, of records need to be processed in a short (and predictable) amount of time, this economy can make all the difference. (Even a mechanical operation that takes "just 1/1000 of a second" to complete, if repeated 10 million times, adds more than two hours to the processing time! Furthermore, the reasons for this delay are purely mechanical, regardless of the computational speed of the computer, and they are a severe bottleneck to the entire process.) Guide to the unit record equipment articles See also Notes | |||||||||
|
| ||||||||||
 |
|
| |