| Most of the research regarding distributed real-time computing has not dealt with the situation in which distributed processes are in competition for overlapping sets of resources. However, it has been recognized that when tasks must compete for mutually exclusive access to overlapping sets of resources, that chains of blocking of indeterminate length can occur. Where tasks are allowed to block and wait on each other, particularly where overlapping resource requirements are concerned, the worst case wait for a task's critical section becomes less well bounded. Moreover, it has been recognized, that due to the tractability of the complexity, resource scheduling that encompasses more than CPU scheduling does not lend itself well to mathematical analysis. In terms of distributed scheduling, some research has been done, but very little has touched upon the priority scheduling of tasks as it relates to real-time computing. In terms of research aimed at investigating real-time resource scheduling in the distributed case, little has been done. Even where tasks run on different CPU's in a distributed system, it has been a general underlying assumption of real-time research, that tasks are independent of each other and do not share physical resources. For many applications, this assumption does not hold.;The principle aim of this research is to design distributed resource scheduling algorithms which allow resource scheduling under real-time constraints and to analyze their performance. A large portion of this thesis has involved the design and case wait for a task's critical section becomes less well bounded. Moreover, it has been recognized, that due to the tractability of the complexity, resource scheduling that encompasses more than CPU scheduling does not lend itself well to mathematical analysis. In terms of distributed scheduling, some research has been done, but very little has touched upon the priority scheduling of tasks as it relates to real-time computing. In terms of research aimed at investigating real-time resource scheduling in the distributed case, little has been done. Even where tasks run on different CPU's in a distributed system, it has been a general underlying assumption of real-time research, that tasks are independent of each other and do not share physical resources. For many applications, this assumption does not hold.;The principle aim of this research is to design distributed resource scheduling algorithms which allow resource scheduling under real-time constraints and to analyze their performance. A large portion of this thesis has involved the design and verification of various novel resource scheduling algorithms. Experiments are conducted in which various aspects of the performance of these protocols are investigated. |
