Online Traffic Prediction in the Cloud: A Dynamic Window Approach

Traffic prediction is a fundamental tool that captures the inherent behavior of a network and can be used for monitoring and managing network traffic. Online traffic prediction is usually performed based on large historical data used in training algorithms. This may not be suitable to highly volatile environments, such as cloud computing, where the coupling between observations decreases quickly with time. We propose a dynamic window size approach for traffic prediction that can be incorporated with different traffic predictions mechanisms, making them suitable to online traffic prediction by adapting the amount of traffic that must be analyzed in accordance to the variability of data traffic. The evaluation of the proposed solution is performed for several prediction mechanisms by assessing the Normalized Mean Square Error and Mean Absolute Percent Error of predicted values over observed values from a real cloud computing data set, collected by monitoring the utilization of Dropbox.