Main Trend Extraction of the Storage Volume for Internet Data Center

For a data center, it is very valuable to predict its storage volume value. According to the trend of current and previous storage volume data, we can predict its future value. However, the real storage volume series is always "dirty", which means it contains noise, missing data and outliers. Hence extracting its main trend is necessary for making an accurate prediction. Otherwise, the "dirty" information will distort the real condition of the data center. In this paper, we propose an estimation method to accurately extract the main trend of storage volume time series. The Kalman filter is used to estimate a compressed storage volume series through irregular sampling method and then the Cubic Spline Interpolation approach is applied to reconstruct the main trend. By applying the Max/Min/Mean average method, we can give a steady main trend. Experiments show that the proposed method can accurately estimate the main trend of the storage volume series, which is helpful to make accurate prediction.