Dynamic Sample Weighting to Predict the Remaining Useful Life of Hard Disk Drives

• Gabriel S. Felix Universidade Federal do Ceará (UFC)
• Francisco F. Pereira Universidade Federal do Ceará (UFC)
• Francisco D. Praciano Universidade Federal do Ceará (UFC)
• João P. Gomes Universidade Federal do Ceará (UFC)
• Javam C. Machado Universidade Federal do Ceará (UFC)

DOI: https://doi.org/10.5753/kdmile.2023.232905

Resumo

Hard Disk Drives (HDDs) are widely used for data storage in various applications. However, their failure can result in significant data loss and system downtime. Therefore, accurate prediction of the remaining useful life (RUL) of HDDs is crucial for proactive maintenance and data backup strategies. In this paper, we propose a novel approach to predict the RUL of HDDs using Long Short-Term Memory (LSTM) networks and incorporating weighted loss functions. The proposed model leverages the Self-Monitoring, Analysis, and Reporting Technology (SMART) features of HDDs, which provide valuable information about the health of the drive. We evaluated two weighting approaches that improve the general performance and enhance predictions within a given timeframe. Our experiments showed that the models outperformed traditional methods in terms of Mean Squared Error (MSE) at given time intervals.

Palavras-chave: HDD, RUL, Failure prediction, Deep Learning, Sample Weighting

Referências

Backblaze. Hard drive data and stats. [link], 2023. Accessed: 2023-02-13.

Cahyadi and Forshaw, M. Hard disk failure prediction on highly imbalanced data using lstm network. In 2021 IEEE International Conference on Big Data (Big Data). pp. 3985–3991, 2021.

Chaves, I. C., de Paula, M. R. P., Leite, L. G., Queiroz, L. P., Gomes, J. P. P., and Machado, J. C. Banhfap: A bayesian network based failure prediction approach for hard disk drives. In Intelligent Systems (BRACIS), 2016 5th Brazilian Conference on. IEEE, pp. 427–432, 2016.

Hochreiter, S. and Schmidhuber, J. Long short-term memory. Neural computation 9 (8): 1735–1780, 1997.

Hu, L., Han, L., Xu, Z., Jiang, T., and Qi, H. A disk failure prediction method based on lstm network due to its individual specificity. Procedia Computer Science vol. 176, pp. 791–799, 2020. Knowledge-Based and Intelligent Information & Engineering Systems: Proceedings of the 24th International Conference KES2020.

Lima, F. D. S., Pereira, F. L. F., Chaves, I. C., Gomes, J. P. P., and Machado, J. C. Evaluation of recurrent neural networks for hard disk drives failure prediction. In 2018 7th Brazilian Conference on Intelligent Systems (BRACIS). IEEE, pp. 85–90, 2018.

Lima, F. D. S., Pereira, F. L. F., Chaves, I. C., Machado, J. C., and Gomes, J. P. P. Predicting the health degree of hard disk drives with asymmetric and ordinal deep neural models. IEEE Transactions on Computers 70 (2): 188–198, 2021.

Murray, J. F., Hughes, G. F., and Kreutz-Delgado, K. Machine learning methods for predicting failures in hard drives: A multiple-instance application. J. Mach. Learn. Res. vol. 6, pp. 783–816, 2005.

Olah, C. Understanding lstm networks, 2015. [Online; accessed 2017-04-26].

Ottem, E. and Plummer, J. Playing it smart: The emergence of reliability prediction technology. Tech. rep., Technical report, Seagate Technology Paper, 1995.

Pereira, F. L. F., Bucar, R. C. B., Brito, F. T., Gomes, J. a. P. P., and Machado, J. C. Predicting failures in hdds with deep nn and irregularly-sampled data. In Intelligent Systems: 11th Brazilian Conference, BRACIS 2022, Campinas, Brazil, November 28 – December 1, 2022, Proceedings, Part II. Springer-Verlag, Berlin, Heidelberg, pp. 196–209, 2022.

Pinheiro, E., Weber, W.-D., and Barroso, L. A. Failure trends in a large disk drive population. In 5th USENIX Conference on File and Storage Technologies (FAST 07). USENIX Association, San Jose, CA, 2007.