An application oriented evaluation method for network performance reliability

Abstract

Congestion phenomenon has been observed in communication networks for many years, and one important reason for this is the effect of network applications. In actual networks, different applications may be provided by the same server node. This results in coupling between the applications. This coupling relationship makes it difficult to evaluate actual network reliability, with network applications. So an effective method to evaluate the network reliability considering the actual use of network applications is required. In this paper, we focus on the evaluation of network performance reliability with the consideration of two different network applications. First, the actual use of two network applications are divided into three different situations: DCA (double coupled applications), SCA (single coupled applications) and NCA (no coupled applications) respectively. Then, considering the network applications, we improve the traffic model, which is commonly used to analyze the dynamics of traffic in complex network. Finally, based on the improved traffic model, an application oriented evaluation method is proposed to evaluate the network performance reliability considering the user's tolerance thresholds. The coupling coefficient is used to measure the coupling strength between two applications. The network performance reliability on a lattice network adopting the shortest path routing strategy with the DCA, SCA and NCA are studied respectively and some interesting results are observed. When the number of application server nodes is small enough compared with the network scale, coupled nodes in the network have significant impact on the network performance reliability and congestion is more likely to happen in coupled no des. When the number of the server nodes is fixed, the network is more reliable with the NCA and more unreliable with the DCA. With the same coupling coefficient, the network is more reliable with the DCA than that with the SCA. With the SCA and the same coupling coefficient, the smaller the bias from the average applications scale is, the more reliable the network is.