An improved parallel banker's algorithm based on Petri net

This thesis explores the serial detection process of traditional banker's algorithm which can be parallelized through the synchronization and sharing synthesis of the Petri Net, and also it puts forward an improved banker's algorithm which is able to carry out parallel deadlock detection on the multi-core platform. First, it analyses the parallel performance of intra-processes and inter-processes so that their state changes and transition sequences can be grasped. Second, it calculates the performing sequences of the parallel algorithm with distributing them on every core of the processor to shorten the time of deadlock detection. The experiment result indicates that such an improved algorithm can greatly raise the detection efficiency and improve multi-core computer resources utilization effectually.