In cloud environment, multiple resources performance prediction is the task of predicting different resources by considering the differences from multiple task inferences based on the historical values to make effective and certainty judgmental decisions for the future values. One resource performance prediction can conclude the performance of another, which implies dependency (i.e., multi-resources) or independency (i.e., one resource), but that cannot be directly confirmed accurately. We use time series algorithms to investigate possible approaches, which can greatly assist us to analyze and predict the future values based on previously observed values. The goal of this paper is to review the theory of the common several models of multivariate time series, and to emphasize the practical steps to take in order to fit those models to real data and evaluate the outcome. Moreover, ensemble-learning algorithms are applied to the best-fit models to improve performance. Finally, we will discuss the results.