Description, modeling and simulation of the congested link of a university

Abstract This paper presents a performance evaluation study of a congested communication network in a university. Queuing often occurs in this network due to high demand and disposal of requests at peak times, resulting directly in decreased service quality for its users, who have their connections delayed or interrupted and cannot access the web pages. The goal of the present work was to characterize this congestion problem empirically, capturing packet traces in logs and studying delays and losses, and to model and analyze this system based on the queuing theory and discrete event simulation to evaluate its performance, identify bottlenecks and proposing solutions to improve the quality of service. The study showed that, in situations of peak traffic, simulation was the quantitative approach that produced results closer to the empirical ones, as the analytical queuing models studied did not provide good approximations, mostly by not reflecting the control exercised by the Transmission Control Protocol (TCP).