Failure Prediction Mechanism of Disk Devices Based on LSTM
Zhenpeng Xu, Jinwei Ma, Yu Liu · 2022 2nd Asia-Pacific Conference on Communications Technology and Computer Science (ACCTCS) · 2022
In order to reduce the impact of the disk device failure on the information system performance fluctuation and business services, a disk device failure prediction mechanism based on LSTM is proposed based on the disk state feature extraction and neural network model. Firstly, based on the SMART information, the performance data and the function attenuation characteristics of the disk, the characteristic index of the disk device is selected and extracted. Then, a disk device failure prediction mechanism is established based on the improved LSTM by using the characteristic index as the training set. Finally, the active failure prediction of disk device was developed. Performance analysis shows the failure prediction mechanism of disk device proposed can obtain online prediction function through the data pre-processing methods such as SMART data item selecting, disk loss feature construction, the data item cleaning, and the sample standardization. The proposal is superior to the traditional neural network model in terms of the error square sum and correction decision coefficient.