Communication Network Modeling And Performance Analysis Based On Tandem Queueing Network Theory

With the rapid development of wideband wireless access technology and mobile communication technology, various wireless access networks with unique advantages emerge one after another. Most experts and scholars agree that network convergence is the direction of network evolution. The existing and arising various wireless access networks will coverage all areas where some hot areas may be overlapped by several networks and finally construct a seamless connection heterogeneous network based on IP technology, forming ubiquitous network. Internet of things has drawn worldwide attention ever since it is put forward about ten years ago. Especially in the recent two years, Internet of things is being researched and deployed vigorously and service target is being extended gradually from human to machine. It can be foreseen that in the near future any object even a non-IT device can communicate, forming ubiquitous service. It is obvious that this ubiquitous communication network will change people's life significantly.Network modeling and performance analysis are important theoretical basis for study, design and deployment of communication network. As a mathematical tool based on stochastic process theory, queueing theory is often used to model communication network and analyze its performance. However, the traditional single node queueing model is inappropriate to model the complex communication network. In addition, network operators and subscribers care more about the end-to-end system performance metrics. So in order to model and analyze the entire communication network, this paper proposes several analytical models based on queueing network theory to analyze the ubiquitous communication networks.Firstly, we model and analyze the performance of M2M service carrier network. Since most M2M service transmit packets frequently with small data size, resulting M2M terminals transform frequently between active and inactive states, which makes M2M traffic is appropriate to be modeled as IBP. In the actual communication scenario, the generated packets are transmitted from one node to another. Therefore, the entire communication network is suitable be modeled as a discrete-time tandem queueing network. Furthermore, we add RED congestion control mechanism in each node of the tandem queueing network because the substantial M2M terminals may cause network congestion. A decomposition method is used to solve the discrete-time tandem queueing network with RED mechanism. We decompose the proposed tandem queueing network into several independent sub-networks, analyze the departure process of each sub-network and then fit the departure process with another IBP model that is used as the input process of next sub-network. In this manner, we can sequentially solve the rest sub-networks and analyze how traffic burstiness and parameters of RED mechanism can affect the performance metrics.Secondly, we also model the communication network carrying H2H service as a discrete-time tandem queueing network with RED mechanism and the H2H traffic is modeled as a MMBP-2 model since it can capture the properties of both burstiness and correlations while still maintaining analytical tractability. We utilize a more complicated departure process analysis and parameter fitting scheme to solve the tandem queueing network because at least four statistics are needed to describe a MMBP-2 model. We investigate the effect of traffic correlations on performance metrics and the effect of parameter setting of RED mechanism on the end-to-end performance metrics.Finally, in order to study on the bearing capacity of communication network that will carry M2M service and H2H service simultaneously and analyze how the two kinds of traffic with different characteristics will effect each other, we model and analyze the communication network with mixed traffic as the input process. According to superposition property of D-BMAP, the entire communication network can be modeled as an aggregate discrete-time tandem queueing network. Using the four statistics derived from departure process analysis, we can approximately fit the departure process of sub-network into a MMBP-2 model. Besides, we build a simplified model of discrete-time tandem queueing network with mixed traffic. The modification of server is to reflect the interference of M2M traffic. By solving the two analytical models, we can obtain various performance metrics that help us find the bottlenecks of the communication network and take measures to improve network performance.