A large semaphore based operating system
The paper describes the internal structure of a large operating system as a set of cooperating sequential processes. The processes synchronize by means of semaphores and extended semaphores (queue semaphores). The number of parallel processes is carefully justified, and the various semaphore constructions are explained. The system is proved to be free of “deadly embrace” (deadlock). The design principle is an alternative to Dijkstra's hierarchical structuring of operating systems. The project management and the performance are discussed, too. The operating system is the first large one using the RC 4000 multiprogramming system.