Análisis de demanda de ancho de banda en los servicios de internet de RedUNAM

Abstract

The growing demand for network services  has raised their importance because they support organizations key processes, which creates the need  for a reliable future perspective for the contracting and renewing internet services. This scenario and the emergence of different models for data study and forecasting that allow the analysis of future behavior using information with time series data, has favored the implementation of strategies that deliver results in less time. Following the  model of continuous improvement and good practices in information technologies services, the Departamento de Monitoreo de la Red, known as NOC RedUNAM, proposed the analysis of behavior of the collected data regarding the bandwidth consumption of internet links that integrate RedUNAM, which exhibit time series characteristics. With the aim of obtaining reliable information about future expected internet service consumption, different analysis algorithms that consider the conditions that affect the demand of services in the Universidad Nacional Autónoma de México (schedules, extraordinary events, school periods, among others) were tested. Based on this need, the algorithm that best  accepted the influence of changes in data links daily behavior was selected. Consequently, the analysis was focused on algorithms that gave major importance to the variables considered relevant for the demand of network links forecast, as well as the network operation general behavior.  For this exercise, simple regression models, integrated moving averages and other techniques were tested, concluding with the selection of an algorithm that offered a margin of reliability, allowing the forecast to be considered acceptable for the conditions of the collected network data.