The UNIVAC File Computer
Unisys History Newsletter.
Volume 2, Number 2
December, 1993 (revised 1999)
by George Gray
The UNIVAC File Computer was sometimes regarded as a rival which delayed the introduction of the UNIVAC Solid State computer, but it would be unfortunate for the File Computer to be remembered only in that way, because it was a very innovative machine. The computer historian Saul Rosen has said that it was perhaps a little bit too innovative, since problems in getting the File Computer to work caused its deliveries to be late, adding to a reputation for lateness which Remington Rand developed during the 1950s.
Although Remington Rand had its own data processing research and development effort in Norwalk, Connecticut, it became the leading (and for a brief time, the only) computer vendor in the world through its acquisitions of the Eckert-Mauchly Computer Corporation (EMCC) in February 1950 and Engineering Research Associates (ERA) in December 1951. These two organizations and the Norwalk group continued to function with a high degree of autonomy, indeed rivalry, for several years. Not until 1955 was there a manager (William Norris from ERA) in charge of all three groups. EMCC had developed the original UNIVAC computer, later referred to as the UNIVAC I when other machines were given the UNIVAC name, while ERA had built its 1101 computer for use by the National Security Agency (NSA). At the time of its acquisition by Remington Rand, ERA was working on a successor to the 1101, which was called the 1103 and was first delivered in October 1953.
Both the UNIVAC I and the 1103 were large and expensive machines, costing around $1 million. IBM entered the large computer market in May 1952 with its announcement of the IBM 701 in the same price range, which was first delivered in April 1953. Many companies, government agencies, and universities could not afford these large computers. Magnetic drum memory was slower and cheaper than the mercury memory of the UNIVAC I and the electrostatic memory of the UNIVAC 1103 and IBM 701. It was possible, then, to make a smaller, slower computer for a lower price. IBM entered this market with its 650 computer, announced in July 1953 and delivered in 1954, which had a drum memory of 2,000 10-digit words. Selling for $200,000 to $400,000 it became a great success: eventually over 1,800 were sold or leased. In 1954, Electrodata Corporation, which was acquired by Burroughs two years later, also brought out a drum computer called the DATATRON.
Unless it was going to give up the medium-scale computer market to IBM and Electrodata, Remington Rand had to respond with a product of its own. The ERA division, which was located in St. Paul, Minnesota, had extensive experience with drum memory computers. It had developed a computer for the Civil Aeronautics Administration to store and retrieve flight plans and another for the U.S. Navy's Logistics Research Project at George Washington University. ERA had also built an on-line interactive computer for inventory control called the Speed Tally System for the John Plain Mail Order Company of Chicago. It was delivered in 1954, and Remington Rand decided to market an improved version of it as a medium-scale computer to be called the UNIVAC File Computer. Development began in January 1955 as a joint effort by the St. Paul group and the tabulating machine laboratory in Norwalk, Connecticut.
As its name indicated, the File Computer was intended to provide access to data files stored on magnetic drums. This was very unusual for that time. Most computers could read in data records from punched cards, paper tape, or magnetic tape, process them, and write them back out to cards or tape, but they had no provision for long-term on-line storage of data. The only comparable computer was IBM's RAMAC, first delivered in 1957, which was a pioneer in the development of disk drives.
The original version of the File Computer, which was called the File 0, had no stored program capability, so the program had to be set up by wiring a plugboard. There was a 1,070-word drum memory, which had twelve 6-bit digits or characters per word. The computer performed decimal arithmetic on numbers represented in Excess-3 (XS-3) code. In XS-3, each decimal digit was represented by a binary value three higher, that is, 0 by 000011, 1 by 000100, etc. There were also codes for letters and special characters, such as 010100 for A, 010101 for B, and so on. The 1,070-word drum, which had an average access time of 2.5 milliseconds, was for storage of data actually being worked on by the program. General data storage was provided by from one to ten 15,000-word drums, whose average access time was 17 milliseconds. It was also possible to have a processor which did tape sorting and collating, separately from the central processor.
Besides doing the usual sort of batch work, the File Computer could provide interactive access to the stored data via one to ten terminals, which were called "input-output devices." An input/output device consisted of an inquiry typewriter, a punched card unit (for either 80 or 90 column cards), a paper tape unit, a printer, and optionally a magnetic tape drive. At least one customer, the Clark Equipment Company, chose the File Computer over an IBM 650 because it was faster at reading paper tape. The input/output devices could operate independently from the central processing complex and perform input/output operations, while the processor was doing something else.
As we have seen, the File Computer had a complex design, what with the high-speed drum, the general storage drums, the tape collating processor, and the input/output devices. The first File 0 was delivered to Douglas Aircraft in Long Beach, California sometime in 1956 to be used for inventory control on the DC-8 aircraft project. The fact that the File 0 could only be programmed by plugboard was a serious limitation, and the engineers in St. Paul worked on providing core memory and an internal programming capability. This version, called the File 1, encountered serious development problems, and missed many promised completion dates. In the meantime, thirty File 0 models were built and shipped to customers.
The first File 1 was built in 1957, but quantity deliveries did not begin until the autumn of 1958. The File 1 had twenty words of core storage, a tiny amount, but enough to provide for some internal programming capability. It allowed for a combination of internal and plugboard instructions in the same program. There were 27 instructions in the internal instruction set, including the arithmetic operations, comparison, jumps, suppression of left zeros, test for input from i/o station, and transfer of control to the plugboard. The plugboard had 19 operations, one of which was transfer of control to the program in memory. The memory instructions were twelve decimal digits in length: the last three specified the instruction code, while the others comprised three three-digit storage addresses. Most File 0 computers were field upgraded to add the core memory and become File 1s. Somewhat later another field upgrade, which replaced the main memory drum with 1740 words of core memory, made the File 1 into a File 2, increasing the operating speed by a factor of 3.6.
Because of the delivery delays, File customers had to be patient. Clark Equipment Company sent two programmers to St. Paul in 1956 to learn programming on the File 0, so that they could start writing applications for a File 1 which was supposed to be delivered in September 1957. Since there were no higher-level programming languages available for it (or for any computers aside from the UNIVAC I and the IBM 701), they probably needed that length of time to work on their payroll and inventory management systems. Eastern Airlines used the File Computer for its reservations system. During 1957, Eastern and the St. Paul engineering staff developed a terminal device, called an agent set, for use on the File Computer, which had a keyboard and push-button matrix for selection of the function to be performed. A demonstration of the agent set took place in August 1957, but the reservation system did not start running until September 1958. The system covered nine cities (Atlanta, Birmingham, Boston, Houston, Miami, New Orleans, New York-Newark, Philadelphia, and Washington) and within two months of the start-up was processing one transaction per second. Northwest Airlines started using a File Computer for its reservations system in November 1959 and Capitol airlines followed in early 1960. It should be noted that these were seat inventory systems, which tracked seats sold versus seats available, not the type of passenger name reservation systems which were developed in the late 1960s.
Douglas Aircraft Company was probably the biggest customer for the File, with a File 0 computer and six File 1s in southern California and another File 0 in Tulsa, Oklahoma. The File 1s were installed between August 1958 and March 1960. They were used for production scheduling, parts inventory, general accounting, time reporting, and payroll. The number of general storage drums on the Douglas machines varied from one to eight. Other users of the File computer included the Chesapeake and Potomac Telephone Company, Michigan Bell, First National City Bank, Western Electric, the Army, the Air Force, and the Marine Corps. These sites used the File primarily for inventory, accounting, and payroll. A basic system with four tape drives, the high-speed drum, and one general storage drum cost about $560,000, while additional drums cost about $20,000 each. For computers which were rented, the monthly charges (including maintenance) varied from $10,000 to $20,000 per month, depending on the number of drums. Somewhere around 180 File Computers were built. Apparently this number was not enough to cover the development costs.
Users of the File Computer liked the ability to have relatively large quantities of data for on-line access and praised the off-line tape sorting and collating capabilities. The small amount of core memory was a severe limitation, and probably hurt sales more than did the existence of the Solid State computer. Because of its size, complexity, and price, the File was not really a competitor for the IBM 650: the File was really suited for bigger applications. The Solid State was more nearly comparable to the 650 in price and capability, and the length of time it took to get it from prototype to the market was unfortunate.
The development delays meant that the File Computers were delivered toward the end of the vacuum-tube (i.e., first) generation of computer hardware. Transistor (second generation) computers began appearing in quantity in 1960 and 1961. They made vacuum-tube machines such as the File Computer obsolete. UNIVAC did produce transistor computers with generally similar characteristics, namely the 490 and the 418 families, but there was no easy migration path. Programs would have needed total rewrites. Eastern Airlines and Northwest Airlines did switch to 490s for their reservations systems in 1962 and 1965 respectively. The File Computers at Yamaichi Securities in Japan and the three Marine Corps Supply Centers were replaced with UNIVAC IIIs in 1963 and 1964. Clark Equipment Corporation replaced its File with a UNIVAC 1108, leaping straight from a first generation to a third generation computer.
The File Computer is an example of Remington Rand's tendency to bite off more than it could chew. This pattern of late delivery of very ambitious designs continued with the LARC and the 1107, resulting in significant financial losses and a declining share of the computer market. The revenues from the smaller Solid State and 1004 computers kept the company afloat in the early 1960s, and the success of the 1108 returned it to profitability and increased market share.
Unisys and UNIVAC are registered trademarks of Unisys Corporation. Copyright 1993, 1999 by George Gray