

This work describes a novel approach to Web performance research. Knowledge Engineering was relating to spatio-temporal Web performance prediction with Sequential Gaussian Simulation method, one of the geostatistical simulation methods. A great advantage of these methods is the possibility to make area-time forecasts. What is more, these methods require only the minimum amount of input data, however they still take into account the geographical location of Web servers and the download time of a given resource. The data for the research were used from an active experiment conducted by our multi-agent measurement system MWING. This system is composed of agents whose main task is to monitor a defined group of Web servers. In the research, the connections between an agent in Gdansk and European servers were considered. First, preliminary analysis of measurement data was carried out, and then structural analisis of data was performed. Both of these analyses are necessary to perform a forecast using Sequential Gaussian Simulation method. Subsequent step was a spatial forecast of a total time of downloading data from Web servers with one-week time in advance. Results of forecast were analysed in detail indicating future directions of research to improve Web performance forecasts.