Uso de técnicas de detecção de comunidades para análise de redes ópticas

Abstract

The growth in the use of services on the Internet has promoted a increasing demand for high transmission rates. This demand have been met by optical networks. At the design stage of these networks, the engineer must be able to assess the performance of a given network before its actual physical implementation. In this design process, several topologies are considered. The comparison between topologies is made through metrics that indicate a certain aspect of the network. Typically the metrics considered are performance indicators, such as Throughput, Blocking Probability, Resilience and also other indicators, such as the network Cost. Performance indicators are important because they inform about the quality of a particular topology. Therefore, performance metrics are essential for the design projects of such networks. The most reliable way to calculate the values of these performance indicators is through simulations. However, simulations have a high computational cost, increasing the time needed to obtain information about topologies; since, in these projects, a very large number of different topologies must be considered. On the other hand, a large number of researches in the most diverse domains of knowledge have been carried out on the theme of community detection in graphs. However, there are no applications of these techniques in high capacity fiber-optic networks. Thus, the present work aims to investigate the existence of a correlation between the ability of a fiber-optic network to form communities and its performance indicators. More specifically, it’s Blocking Probability and indicators of Resilience. The analysis was performed comparing the Blocking Probability and Resilience of these networks and the clustering metrics using scatter plots. According to the results, there is a positive correlation between the community metrics and the network performance indicators, and comparatively a speedup of approximately 4,500 times was obtained between the community metrics and the simulations.