Storage Device Performance Prediction with SBCART (Selective Bagging Classification and Regression Tree) Models

Storage device performance prediction is a key element of self-managed storage systems and application planning tasks, such as data assignment and configuration. Based on bagging ensemble, we proposed an algorithm named selective bagging classification and regression tree (SBCART) to model storage device performance. It can improve the accuracy of a single CART model. In contrast with CART modeling, the black-box SBCART model predicts the performance with the higher accuracy. Experiments indicate that SBCART used in storage device performance prediction is effective and stable. In addition, we use the caching effect as a measure in feature vector to make good predictions.