Deep Learning-based Real-time Failure Detection of storage devices

• Main Authors: Lien-Chung Tsai, 蔡蓮忠
• Other Authors: Chuan-Jun Su
• Format: Others
• Language: en_US
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/j46n3w

碩士 === 元智大學 === 工業工程與管理學系 === 106 === With the rapid development of cloud technologies, evaluating cloud-based services has emerged as a critical consideration for data center storage system reliability, and ensuring such reliability is the primary priority for such centers. Therefore, a mechanism by which data centers can automatically monitor and perform predictive maintenance to prevent hard disk failures can effectively improve the reliability of cloud services. Predictive maintenance differs from traditional preventative maintenance or repair in that it allows for early detection of potential failure on currently operating equipment. This study develops an alarm system for self-monitoring hard drives that provides fault prediction for hard disk failure. Combined with big data analysis and deep learning technologies, machine fault pre-diagnosis technology is used as the starting point for fault warning, and is combined with self-monitoring, analysis and reporting technologies (SMART) to identify abnormal operations before failure. Finally, a predictive model is constructed using Long and Short Term Memory (LSTM) Neural Networks for Recurrent Neural Networks (RNN). The resulting monitoring process provides condition monitoring and fault diagnosis for equipment which can diagnose abnormalities before failure, thus ensuring optimal equipment operation.