Host Load Prediction with Bi-directional Long Short-Term Memory in Cloud Computing

Abstract

Host load prediction is the basic decision information for managing the computing resources usage on the cloud platform, its accuracy is critical for achieving the servicelevel agreement. Host load data in cloud environment is more high volatility and noise compared to that of grid computing, traditional data-driven methods tend to have low predictive accuracy when dealing with host load of cloud computing, Thus, we have proposed a host load prediction method based on Bidirectional Long Short-Term Memory (BiLSTM) in this paper. Our BiLSTM-based apporach improve the memory capbility and nonlinear modeling ability of LSTM and LSTM Encoder-Decoder (LSTM-ED), which is used in the recent previous work, In order to evaluate our approach, we have conducted experiments using a 1-month trace of a Google data centre with more than twelve thousand machines. our BiLSTM-based approach successfully achieves higher accuracy than other previous models, including the recent LSTM one and LSTM-ED one.