Predictive Maintenance of Storage Devices Using Deep Learning Techniques

Abstract

The requirement of storage devices is of great significance in several organizations. As the data is increasing at a tremendous pace so is the requirement to store and handle this data securely. To store this data we make use of storage devices such as hard disk drive or solid state drive. These storage devices also are equipped with a SMART(self monitoring and reporting technology) system that monitors the various parameters of the storage device for example temperature. The deep learning models such as the LSTM(Long Short Term Memory) network can be used to predict whether the storage device will be experiencing a failure or not in the given future. In this study the LSTM model is compared with the RNN(Recurrent Neural Network) and GRU(Gated Recurrent Neural Network) models, which concluded that the LSTM gave a better performance since it is able to find the long term dependencies in the data. The precision of the LSTM model is 84% which is greater than the other two models. The LSTM model also gave better results in terms of recall, f1 score and accuracy, considering the predictive nature of the binary classification problem.