Transaction-based model for real-time distributed control systems

The paper presents a transaction-based task allocation model suitable for real-time distributed control applications. In accordance with the proposed model the autonomous functionalities of a control application are implemented as chains of tasks and messages, called transactions. This model allows the evaluation of worst case response time for transactions and subsequently determines the real-time feasibility of the system. The task allocation strategy assures fault-tolerance and adaptable real-time behavior. As communication infrastructure a general purpose switched Ethernet network was considered. The transaction-based model was validated through simulations made with a Timed Petri-Net tool.