In this thesis, we consider the performance problem of a hypothetical computer network. We use the term capacity to denote the maximum amount of traffic that a computer network can carry. We require that there are no losses in transmissions and that the travel times of IP packets over the network are acceptably small. We have considered the problem of electricity meters that send their measurements to a server over a backbone network. We briefly present some software environments that allow us to build a network model and run simulations on it. The built model, the settings and the operation of all the components are then presented in detail. We also explain the flow of different types of data over the network. We perform four sets of simulations on the network model: with single-router routers with deterministic distribution of inter-arrival times of primary requests, with multi-router routers with deterministic distribution of inter-arrival times of primary requests, with exponential distribution of inter-arrival times of primary requests, and simulations with a large amount of secondary traffic. In the first set of simulations, we examine the number of lost requests and the length of waiting times at the baseline settings. In the second set of simulations, we get rid of request losses and reduce the waiting times at the routers by increasing the number of routing circuits. In the third set of simulations, we then introduce an exponential distribution of inter-arrival times of primary requests, which is a better approximation of the dynamics in real networks. In the last set of simulations, we observe the network efficiency at higher traffic volumes when the intensity of the arrival of secondary requests is increased.
|
