Time-Sharing Computer Systems | Research & Encyclopedia Articles

Time-Sharing Computer Systems

Time sharing enables large computers to interact with and perform tasks for many users at such a fast rate that users are practically unaware the computer's time is divided. Before the development of time-sharing systems in the 1960s, computers were generally large and expensive and ran only one program at a time. Users took turns on the computer. Running a single program could take hours because it involved manually setting up the whole computer system: computer, tape drives, card punchers and readers, and printers. An iterative computer task, such as debugging a program, could take many lengthy sessions with the computer. During setup, the computer itself sat idle--a waste of an expensive, valuable resource.

Operating systems (OS) and batch processing were early developments that made computers more efficient. Operating systems were developed in the mid-1950s as sets of special computer programs that controlled peripheral devices. While the OS handled the peripherals for one program, the processor moved on to address the next program. Eventually, methods were developed for running sets of several programs in quick succession. This was called "batch processing," and it greatly increased computer efficiency. Batch-processing computers were often managed by special departments within organizations. Computer users submitted their "jobs" to the batch-processing department and returned later to pick up the results. So, while batch processing improved the efficiency of computers, it also put a distance between computers and users. This distance was particularly irksome to programmers for whom direct interaction with computers was the better, faster way to develop, test, and debug programs.

Another advance in computer efficiency was realized in the development of "multiprogramming" in the late 1950s. Multiprogramming makes use of an executive program to run several programs at once and to generally organize the computer's processing functions for maximum efficiency.

In the early 1960s, a group of computer experts centered at the Massachusetts Institute of Technology (MIT) who had worked on early, interactive computers began to seek ways to allow multiple users to interact with computers simultaneously. They built upon advances made in the development of multiprogramming to create the first time-sharing systems.

The first time-sharing system was proposed by John McCarthy of MIT and developed by McCarthy and Herbert Teager, also of MIT. The purpose of this first system was to allow programmers to debug their programs while other users were running production programs or debugging other programs. The system was operational at the MIT Computation Center in early 1960. It was an exciting success, and MIT decided to meet all its computing needs with a large, time-sharing computer system that would be remotely accessible from several locations on the campus.

While planning for a campus-wide time-sharing system was still underway, another group at the MIT Computation Center, led by Fernando Corbato, began developing a new time-sharing system in late 1960. This system, called the Compatible Time-Sharing System (CTSS), was operational in 1961. CTSS marked a step forward in time sharing in that it supported up to 21 simultaneous users by 1963. CTSS stored programs in its main memory. Users requested jobs to run via remote terminals. CTSS ran the programs and sent the results to users at the terminals. As soon as one job was complete, the CTSS processor began work on another, so that users were not aware the system was running other jobs. The final version of CTSS included powerful utilities for editing, testing, and debugging programs.

At Bolt Beranek and Newman (BBN), a Cambridge, Massachusetts, company with strong ties to MIT, Edward Fredkin developed a time-sharing system on a smaller computer than those used at MIT. The BBN system included the innovation of dividing the computer's memory into short-term and long-term and of achieving time-sharing by swapping users in and out of short-term and long-term memory. J. C. R. Licklider, another employee at BBN, worked on this system and became a time-sharing enthusiast. Licklider later became the first director of the Information Processing Techniques Office (IPTO) of the Department of Defense's Defense Advanced Research Projects Agency (DARPA). Licklider made research and development in time sharing the primary focus of IPTO when it opened in 1961. With the funding it provided companies and universities to research and develop more advanced time-sharing systems, IPTO became the most important proponent of time sharing.

One of Licklider's first actions as director of IPTO was to alter the only current project--a command and control project with Systems Development Corporation (SDC) of Santa Monica, California--to a time-sharing project. Later, Licklider approved IPTO funding for other time-sharing projects on the west coast, at the University of California at Los Angeles, University of California at Berkeley, the Stanford Research Institute, and Stanford University. Researchers for these projects were able to connect to the SDC time-sharing system when it was completed in 1963.

Licklider also sought to increase existing interest in time sharing on the east coast, specifically at MIT and BBN. With encouragement from Licklider, Robert Fano of MIT received IPTO funding for an MIT time-sharing project. He and Licklider called it Project MAC, for "Multiple Access Computer" and "Machine Aided Cognition." The Project MAC system was operational in 1963. It was based on the CTSS system but quickly outpaced that system.

In 1964 leaders of Project MAC began to plan significant improvements to the system. They contracted with General Electric (GE) and Bell Laboratories to build a new system using a GE computer and Bell software. This new project was called MULTICS, for "Multiplexed Information and Computing Service." MULTICS was to be a cutting-edge system that advanced time-sharing research and development to the next level, while providing MIT with the very best computing services available. However, achieving the next level of time sharing proved to be much harder than anyone had anticipated. MULTICS was expected to be complete in 1965; it was operational--but not complete--by 1969. Interest in and work on MULTICS gradually faded, and the project was never completed.

At the time the MULTICS project began, the major computer manufacturers, including IBM and DEC, offered time-sharing computers and large, mature time-sharing systems were in use in commercial as well as research settings. The promised benefits of time sharing, however, were increasingly viewed with skepticism. Some argued that batch-processing systems were more reliable and more economical than time-sharing systems, and, when designed with enough computing power, could provide turnaround times that were acceptable to programmers as well as general users. It was not batch-processing systems, however, that overtook time-sharing systems. It was the increasingly available small, cheap, yet powerful computer that ended the dominance of time-sharing systems in computing research, development, commercialization, and use.

Although time-sharing systems did not become the dominant method of computer use, they are still used in some settings and they have influenced the nature of current computer use. Time-sharing systems are still used in situations where there is a large computer or peripheral device that needs to be shared by multiple users. And the interactive, shared elements of time-sharing systems have become important, substantial elements in contemporary computer systems.