Multi Agent Systems for Deadlock Prevention

In this book, a shell program acts as an interface among user, agent program and operating system is developed. Aim of the agent program is to handle and solve the deadlock that occurs during the process execution. Since, the memory is one of the most important shared sources; the agent program provides filing operations for memory usage of each running process and replies requests of shell. Learning of the agents are done by using error back propagation learning on Multi-Layer Perceptron feed forward Neural Networks. Initially, shell program requests the process memory usage of running processes, and then it schedules and organizes processes as inputs on the Multi-Layer Perceptron Neural Network. If the output of Neural Network is greater than or equal to a pre-decided threshold shell runs the process. Otherwise, since there exist deadlock occurrence probability, the shell suspends process and prevents the deadlock occurrence. This is a new dynamic approach for deadlock prevention solution. The previous works about this subject is investigated and the test results among exist and proposed solution are concluded.