Autoregressive Time Series Forecasting of Computational Demand

We study the predictive power of autoregressive moving average models when forecasting demand in two shared computational networks, PlanetLab and Tycoon. Demand in these networks is very volatile, and predictive techniqu es to plan usage in advance can improve the performance ob- tained drastically. Our key finding is that a random walk predictor performs best for one-step-ahead forecasts, whereas ARIMA(1,1,0) and adaptive exponential smoothing models perform bet- ter for two and three-step-ahead forecasts. A Monte Carlo bootstrap test is proposed to evaluate the continuous pre- diction performance of different models with arbitrary con - fidence and statistical significance levels. Although the pre- diction results differ between the Tycoon and PlanetLab net - works, we observe very similar overall statistical proper- ties, such as volatility dynamics.