Job Control - History

History

It became obvious to the early computer developers that their fast machines spent most of the time idle because the single program they were executing had to wait while a slow peripheral device completed an essential operation such as reading or writing data. Buffering only provided a partial solution; eventually an output buffer would occupy all available memory or an input buffer would be emptied by the program, and the system would be forced to wait for a relatively slow device to complete an operation.

A more general solution is multitasking. More than one running program, or process, is present in the computer at any given time. If a process is unable to continue, its context can be stored and the computer can start or resume the execution of another process. At first quite unsophisticated and relying on special programming techniques, multitasking soon became automated, and was usually performed by a special process called the scheduler, having the ability to interrupt and resume the execution of other processes. Typically a driver for a peripheral device suspends execution of the current process if the device is unable to complete an operation immediately, and the scheduler places the process on its queue of sleeping jobs. When the peripheral completed the operation the process is re-awakened. Similar suspension and resumption may also apply to inter-process communication, where processes have to communicate with one another in an asynchronous manner but may sometimes have to wait for a reply.

However this low-level scheduling has its drawbacks. A process that seldom needs to interact with peripherals or other processes would simply hog processor resource until it completed or was halted by manual intervention. The result, particularly for interactive systems running tasks that frequently interact with the outside world, is that the system is sluggish and slow to react in a timely manner. This problem is resolved by allocating a "timeslice" to each process, a period of uninterrupted execution after which the scheduler automatically puts it on the sleep queue. Process could be given different priorities, and the scheduler could then allocate varying shares of available execution time to each process on the basis of the assigned priorities.

This system of pre-emptive multitasking forms the basis of most modern job control systems.