Comparing time series with machine learning-based prediction approaches for violation management in cloud SLAs