Since wireless sensor actor networks (WSANs) interact with critical physical environments, one of the important issues of WSANs is real-time considerations. Existing WSANs suffer from the lack of a real-time task allocation in support of real-time communication and coordination. In this paper we present a two level task allocation mechanism. We first break end-to-end periodic tasks into real-time jobs, and then use appropriate algorithms for sensing tasks and acting tasks. To formally state our approach, we propose a model for WSANs using graph transformation systems. Using this formalism we analyze the correctness of our algorithms. We show that the proposed algorithms guarantee that the tasks complete their activities before their deadlines expire. To show the efficiency of our algorithms we have simulated the model. Simulation results showed an improvement of 65 percent in deadline hit ratio comparing our approach to FIFO algorithm.