Spatio-temporal web performance forecasting with sequential Gaussian simulation method

Abstract

In this article, we describe a novel proposal for the spatio-temporal Web performance forecasting using one of the geostatistical methods - Sequential Gaussian Simulation (SGS). Necessary data were obtained from Multiagent Internet Measurement System - MWING, which monitored web transactions issued by MWING's agent located in Las Vegas, US and targeting web servers in Europe. Data contains the measurements, which were taken every day at the same time: at 06:00 a.m., 12:00 a.m. and 6:00 p.m. during the period of May, 2009. First, the preliminary analysis of measurement data was conducted. Next, the structural analysis, which includes directional variogram approximated with the theoretical model, was performed. Subsequently, the spatial forecast (from one week in advance) of total time of downloading data from Web servers was calculated. The analysis of server activity on a particular weekday for a period of a few weeks in selected time intervals and considered forecasted errors was performed. Finally, the results of forecast were analyzed in detail, followed by the determination of subsequent research directions to improve Web performance forecasts. © 2012 Springer-Verlag.