Discrete Event System Modeling and Dynamic Service Allocation for REID Networks

As deployment of RFID readers for tracking and tracing applications in retail stores and manufacturing warehouses becomes more pervasive, data traffic in each reader is expected to grow exponentially in varying load conditions. It is therefore important to ensure that all readers are serviced in a timely manner by RFID middleware server in order to maximize timeliness of information and minimize loss of data. This paper studied the problem of assigning service availability of RFID middleware server to a network of RFID readers. Two methods are evaluated in our simulation study: deterministic and stochastic service disciplines. Deterministic service discipline assigns equal service availability to every reader while stochastic service discipline assigns a higher chance of service availability to readers with heavier data traffic. This paper also presented a discrete event system model of RFID networks for simulation, performance measurement and analysis. One of the simulation results showed that the stochastic service discipline reduces the average response time of readers with heavy data load by 48-73% as compared to the deterministic service discipline. However, the trade-off is poor response time in other readers