Real-time scheduling by parallel and distributed simulation over IP Multicast

Abstract Scheduling optimization of manufacturing processes becomes more and more important. Complex models with a lot of constraints cause a higher computational effort of the simulation. Optimization cycles of heuristic optimization algorithms like genetic algorithms or local search strategies can easily be scheduled in parallel. But recurrent requirements are the distribution of the model, the interchange of the parameters and results. The IP Multicast architecture is an ideal platform for distributing model data over the network. It reduces the network overhead for sending the model to different clients and simplifies also the initial setup between client and server. The client and the server are able to find each other automatically by preconcerted multicast channels. The developed test implementation is designed to work with the simulation system simcron MODELLER and will be used to handle typical optimization tasks like the weekly demand plan of a back-end process or optimal batch sizes for burn-in ovens. It is also suitable for educational purposes and we use it in practical courses where students learn more about factory scheduling. The system offers a high reusability for any operating sequence optimization problem in electronic or semiconductor manufacturing industry, as well as in other fields.