Modelling computer networks

Traffic models in computer networks can be described as a complicated system. These systems show non-linear features and to simulate behaviours of these systems are also difficult. Before implementing network equipments users wants to know capability of their computer network. They do not want the servers to be overloaded during temporary traffic peaks when more requests arrive than the server is designed for. As a starting point for our study a non-linear system model of network traffic is established to exam behaviour of the network planned. The paper presents setting up a non-linear simulation model that helps us to observe dataflow problems of the networks. This simple model captures the relationship between the competing traffic and the input and output dataflow. In this paper, we also focus on measuring the bottleneck of the network, which was defined as the difference between the link capacity and the competing traffic volume on the link that limits end-to-end throughput. We validate the model using measurements on a working network. The results show that the initial model estimates well main behaviours and critical parameters of the network. Based on this study, we propose to develop a new algorithm, which experimentally determines and predict the available parameters of the network modelled.