© 1987 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
Deadlock Prevention in Process Control Computer Systems
1 Department of Management and Information Science, Faculty of Commerce, Hannan University, 5-4-33 Amamihigashi, Matsubara-shi, Osaka 580, Japan, 2 Central Research Laboratories, Engineering Center, Sharp Corporation, 2613-1 Ichinomoto, Tenri-shi, Nara 632, Japan
System deadlock is a serious problem in a multiprogramming environment. The approaches to this problem can be divided into three categories: (1) prevention, (2) detection and recovery, and (3) avoidance. This paper proposes a variation of the first approach, partially applying ideas developed in the second and third approaches. This approach is especially effective in process control computer systems in which the application programs are usually fixed once designed. Using four predetermined application program parameters obtained in the program development stage, a directed graph model and a ‘restriction’ matrix model are introduced representing the usage of common resources. Conditions sufficient for system deadlock prevention are presented along with algorithms for checking to see that the models meet these conditions. By using this approach, if a deadlock possibility is detected the causes can also be detected. The deadlock can thus be prevented during the program development stage. As the algorithms are not used in the real-time mode, there is no negative effect on the responsiveness of the system. A higher utilisation rate of common resources is also ensured because the usage of resources is restricted only when the possibility of a deadlock is detected.
Received June 1985.
* Department of Management and Information Science, Faculty of Commerce, Hannan University, 5-4-33 Amamihigashi, Matsubara-shi, Osaka 580, Japan.
Central Research Laboratories, Engineering Center, Sharp Corporation, 2613-1 Ichinomoto, Tenri-shi, Nara 632, Japan.