The Research Of Differentiated Service Mechanism With Preemptive Priority Strategy In Wmsns

In WMSNs(Wireless Multimedia Sensor Networks), the real-time data (such as audio and video, etc.) has a higher requirement on the transmission delay, and other the non-real-time data (such as temperature of periodic monitoring, etc) has a higher demand on the transmission reliability. DiffServ (Differentiated service) can satisfy these different needs and provide better QoS guarantee for multimedia services. For cluster-based WMSNs, two strategies with preemptive priority mechanism for DiffServ are proposed, the system models are built and the performance are evaluated.Firstly, in order to extremely reduce the delay of the real-time data, and ensure the reliability of the non-real-time data to some extent, we provide the real-time data with non-buffer, and let the real-time data have a higher transmission prior to the non-real-time data. A DiffServ mechanism with preemptive priority and non-buffer denoted as DiffServ_RBN on the real-time data is proposed. Under the condition that the delay requirement of the real-time data is satisfied, in order to improve its transmission reliability to some extent, a certain buffer is set for the real-time data, and the real-time data is given a preemptive priority, we propose a DiffServ mechanism with preemptive priority and finite buffer denoted as DiffServ_RBF on the real-time data.Secondly, the non-real-time data is regarded as a class I customer, the real-time data is regarded as a class II customer. According to the DiffServ_RBN, a discrete-time queueing model Geo₁+Geo₂/Geo₁,Geo₂/1/k+1 with two classes of customers and preemptive priority is built. Moreover, another discrete-time queuing model Geo₁+Geo₂/Geo₁,Geo₂/1/k₁+1,k₂+1 with two classes of customers and preemptive priority is built to capture the working principle of the DiffServ_RBF.Thirdly, by using the method of two-dimensional Markov chain, the transition probability matrixes of the queuing model with non-buffer and finite buffer are given, and the steady state probability distributions of the queueing models are derived. Correspondingly, we give the formulas of the system performance in terms of the average response time, the data dropped ratio, the system throughput and the energy saving ratio for the two DiffServ mechanisms proposed in this paper.Finally, the system performance of the two DiffServ mechanisms is verified by numerical experiments with different system parameters. Considering both the average response time and data dropped ratio, a cost function is constructed to optimize the buffer size.