Early Digital Computing (part 2) 

The AN/FSQ-7 Computer

The Whirlwind project at MIT's Digital Computer Laboratory had demonstrated real-time control, the key ingredient for the Project Lincoln air defense concept. Whirlwind also provided an experimental test bed for the system design. By the spring of 1952, Whirlwind was working well enough to be used as part of the early air defense prototyping activities. The focus of the program within the Digital Computer Division shifted, therefore, to development of a production computer, Whirlwind II.

Whirlwind I was more of a breadboard than a prototype of a computer that could be used in the air defense system. The Whirlwind II group dealt with a wide range of design questions, including whether transistors were ready for large-scale employment (they were not) and whether the magnetic-core memories were ready for exploitation as a system component (they were). The most important goal for Whirlwind II was that there should be no more than a few hours of down time per year.


The AN/FSQ-7 computer was designed by joint Lincoln Laboratory and IBM committees that managed to merge the best elements of their members' diverse backgrounds to produce a result that advanced the state of the art in many directions.
 

To turn the ideas and inventions developed in the Whirlwind program into a reproducible, maintainable operating device required the participation of an industrial contractor. A team, led by Jay Forrester, was set up to evaluate potential contractors.  This team was responsible for finding the most appropriate computer designer and manufacturer to translate the progress made in Project Lincoln into a design for an operational air defense system. The team surveyed the possible engineering and manufacturing candidates and chose three for further evaluation: IBM, Remington Rand, and Raytheon. They visited each company, reviewed its capabilities, and graded each on the basis of personnel, facilities, and experience.

IBM received the highest score and was issued a six-month subcontract in October 1952. Over the next few months, the IBM group visited Lincoln Laboratory frequently to study the system being developed, to become acquainted with the overall design strategy, and to learn the specifics of the central processor design.

In January 1953, system design for Whirlwind II began in earnest. The computer was designed by joint Lincoln Laboratory and IBM committees that managed to merge the best elements of their members' diverse backgrounds to produce a result that advanced the state of the art in many directions.

The schedule was tight. Lincoln Laboratory set a target date of January 1, 1955, to complete the prototype computer and its associated equipment. Installation, testing, and integration of the equipment in the air defense system were scheduled to be started on July 1, 1954. The nine months preceding this, October 1, 1953, to July 1, 1954, would be needed for procurement of materials and construction. The schedule left about nine months for engineering tasks in connection with the preparation of specifications, block-diagram work, development of basic circuit units, special equipment design, and everything else necessary before construction could begin.

In April 1953, IBM received a prime contract to design the computer. A short time later, the name Whirlwind II was dropped in favor of Air Force nomenclature, and the computer was designated AN/FSQ-7. In September, IBM received a contract to build two single-computer prototype systems, AN/FSQ-7(XD-l) and AN/FSQ-7(XD-2). (The XD stands for experimental development.) The AN/FSQ-7(XD-l) replaced Whirlwind in the prototype air defense system during 1955. IBM kept the AN/FSQ-7(XD-2) in Poughkeepsie, New York, and used the machine to support software development and to provide a hardware test bed.

IBM's AN/FSQ-7 computer

The AN/FSQ-7 computer built by IBM was the heart of the SAGE system.

As the plans for the continental air defense system began to take shape, it became evident to the Air Force that automating the combat centers would be desirable. (Each combat center directed operations and allocated weapons for several direction centers.) The combat centers needed a computer like the AN/FSQ-7 but with a specialized display system; this system was named the AN/FSQ-8. The AN/FSQ-8 display console could show the status of an entire sector. Its inputs and outputs did not handle radar or other field data, but were dedicated to communication with direction centers and with higher-level headquarters.

IBM received its first production contract in February 1954. The first AN/FSQ-7 was declared operational at McGuire Air Force Base, New Jersey, on July 1, 1958. IBM eventually manufactured 24 AN/FSQ-7s and 3 AN/FSQ-8s. Each AN/FSQ-7 weighed 250 tons, had a 3000 kW power supply and required 49,000 vacuum tubes. To ensure continuous operation each computer was duplexed; it actually consisted of two machines. The percentage of time that both machines in a system were down for maintenance was 0.043%, or 3.77 hours averaged over a year.

Software Development

In addition to the computer hardware, a large part of the air defense project effort was devoted to software development. The software task developed quickly into the largest real-time control program ever coded, and all the coding had to be done in machine language since higher-order languages did not yet exist. Furthermore, the code had to be assembled, checked, and realistically tested on a one-of-a-kind computer that was a shared test bed for software development, hardware development, demonstrations for visiting officials, and training of the first crew of Air Force operators. 

 

The art of computer programming was essentially invented for SAGE.

Computer software was in an embryonic state at the beginning of the SAGE effort. In fact, the art of computer programming was essentially invented for SAGE. Among the innovations was more efficient programming (both in terms of computer run time and memory utilization), which was achieved through the use of generalized subroutines and which allowed the elimination of a one-to-one correspondence between the functions being carried out and the computer code performing these functions. A new concept, the central service (or bookkeeping) subprogram, was introduced.

Documentation procedures provided a detailed record of system operations and demonstrated the importance of system documentation. Checkout was made immensely faster and easier with utility subprograms that helped locate program errors. These general-purpose subprograms served, in effect, as the first computer compiler. The size of the program—25,000 instructions—was extraordinary for 1955; it was the first large, fully integrated digital computer program developed for any application. Whirlwind was equal to the task: between June and November 1955, the computer operated on a 24-hour, 7-day schedule with 97.8% reliability.

Because of the complexity of the software, Lincoln Laboratory became one of the first institutions to enforce rigid documentation procedures. The software creation process included flow charts, program listings, parameter and assembly test specifications, system and program operating manuals, and operational, program, and coding specifications. About one-quarter of the instructions supported operational air defense missions. The remainder of the code was used to help generate programs, to test systems, to document the process, and to support the managerial and analytic chores essential to good software.

These programs were large, and because MIT did not want further increases in the Lincoln Laboratory staff, the Rand Corporation was asked to assist in the programming task. Rand was eager to play a role in the SAGE effort and began work on the software in April 1955. By December, the section of Rand in charge of SAGE programming—the System Development Division—had a staff of 500.  Within a year, the System Development Division had a staff of a 1000 and was larger than the rest of Rand put together. The Division left its parent company in November 1956 and formed the nonprofit System Development Corporation, with a $20 million contract to continue the work started by Rand and with additional contracts for programming the SAGE computers. SAGE had spawned another company, and another industry.

Back to part 1: Early Digital Computing


Adapted from E.C. Freeman, ed., Technology in the National Interest, Lexington, Mass.: MIT Lincoln Laboratory, 1995.

top of page