An Attention-based GRU Encoder Decoder for Hostload Prediction in a Data Center
Shengpan Qian, Yang Yu, Ling Li, Yanbo Chang · 2021
Data center host load forecasting plays a key role in data center resource scheduling and energy saving. This paper proposes a recurrent neural network encoder and decoder host load prediction model based on the attention mechanism. Using the attention mechanism, the model can more accurately select the factors that affect the current host load. At the same time, through the analysis of the data, it is found that the memory occupancy rate of the host is similar to the host load to a certain extent, so the load of the host can be further accurately predicted by the feature of increasing the memory occupancy rate. The experiment uses a 29-day data center host load data set open sourced by Google. The experimental results show that the model proposed in this paper can achieve the best performance currently.