Attention-Based Fusion of Device Health and Workload Telemetry for Storage-Failure Prediction: An Ablation Study
Abstract
Storage-failure prediction may benefit from combining slow-changing device-health indicators with rapidly changing workload telemetry, but the contribution of each information source is uncertain. We developed an attention-based fusion model using daily device-health features and hourly workload sequences from 8,316 storage devices monitored over 18 months. Performance was compared with late fusion, concatenation, and single-modality baselines using device-separated temporal validation. The full model achieved an AUROC of 0.88 and an area under the precision-recall curve of 0.41, outperforming the strongest single-modality model by 0.07 and 0.09, respectively. Ablation of workload attention reduced early warning sensitivity particularly for controller and thermal events, while removing device-health features had the greatest effect on mechanical-failure detection. Attention weights shifted toward recent queue-depth and latency changes during the final 72 hours before failure. Calibration error remained below 0.04 after temperature scaling. The results indicate that multimodal fusion improves predictive performance because health and workload telemetry capture different precursors. Attention mechanisms were most useful when treated as a fusion component rather than as a substitute for rigorous temporal validation and calibration.
References
1. Nazir M. Performance-telemetry-based anomaly detection for early failure identification in enterprise storage systems. Journal of Multidisciplinary Research and Integrated Sciences. 2023. Available from: https://jmris.com/index.php/jmris/article/view/2023V1P6
2. Khan MFI, Begum MH, Rahman MA, Limon GQ, Azam MA, Masum AKM. A comprehensive review of advances in transformer, GAN, and attention mechanisms: their role in multimodal learning and applications across NLP. International Journal of Science and Research Archive. 2025;15(1):454-459. doi:10.30574/ijsra.2025.15.1.0980.
3. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al. Attention is all you need. Adv Neural Inf Process Syst. 2017;30. Available from: https://arxiv.org/abs/1706.03762
4. Schroeder B, Gibson GA. Disk failures in the real world: what does an MTTF of 1,000,000 hours mean to you?. In: Proceedings of the 5th USENIX Conference on File and Storage Technologies. 2007. p. 1-16. Available from: https://www.usenix.org/legacy/events/fast07/tech/schroeder.html
5. Pinheiro E, Weber WD, Barroso LA. Failure trends in a large disk drive population. In: Proceedings of the 5th USENIX Conference on File and Storage Technologies. 2007. p. 17-29. Available from: https://www.usenix.org/legacy/events/fast07/tech/pinheiro.html
6. Lu S, Luo B, Patel T, Yao Y, Tiwari D, Shi W. Making disk failure predictions SMARTer. In: Proceedings of the 18th USENIX Conference on File and Storage Technologies. 2020. p. 151-167. Available from: https://www.usenix.org/conference/fast20/presentation/lu
7. Murray JF, Hughes GF, Kreutz-Delgado K. Machine learning methods for predicting failures in hard drives: a multiple-instance application. J Mach Learn Res. 2005;6:783-816. Available from: https://www.jmlr.org/papers/v6/murray05a.html
8. Breiman L. Random forests. Mach Learn. 2001;45:5-32. doi:10.1023/A:1010933404324.
9. Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016. p. 785-794. doi:10.1145/2939672.2939785.
10. Guo C, Pleiss G, Sun Y, Weinberger KQ. On calibration of modern neural networks. Proc Mach Learn Res. 2017;70:1321-1330. Available from: https://proceedings.mlr.press/v70/guo17a.html
11. Saito T, Rehmsmeier M. The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLoS One. 2015;10(3):e0118432. doi:10.1371/journal.pone.0118432.
12. Kapoor S, Narayanan A. Leakage and the reproducibility crisis in machine-learning-based science. Patterns. 2023;4(9):100804. doi:10.1016/j.patter.2023.100804.
13. Ovadia Y, Fertig E, Ren J, Nado Z, Sculley D, Nowozin S, et al. Can you trust your model's uncertainty? Evaluating predictive uncertainty under dataset shift. Adv Neural Inf Process Syst. 2019;32. Available from: https://arxiv.org/abs/1906.02530
14. Lakens D. Sample size justification. Collabra Psychol. 2022;8(1):33267. doi:10.1525/collabra.33267.
15. Nosek BA, Ebersole CR, DeHaven AC, Mellor DT. The preregistration revolution. Proc Natl Acad Sci U S A. 2018;115(11):2600-2606. doi:10.1073/pnas.1708274114.
16. Wilkinson MD, Dumontier M, Aalbersberg IJ, Appleton G, Axton M, Baak A, et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data. 2016;3:160018. doi:10.1038/sdata.2016.18.
17. Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863. doi:10.3389/fpsyg.2013.00863.
18. Efron B. Bootstrap methods: another look at the jackknife. Ann Stat. 1979;7(1):1-26. doi:10.1214/aos/1176344552.
Published
Issue
Section
License
Authors retain copyright. Articles published under the Creative Commons Attribution 4.0 International licence (CC BY 4.0) may be shared and adapted for any purpose, including commercially, provided appropriate credit is given, a link to the licence is supplied, and changes are indicated. No additional legal or technological restrictions may be imposed. Third-party material is included only where its credit line permits. Licence: https://creativecommons.org/licenses/by/4.0/. Earlier publications remain subject to their stated licence and author agreements unless the rights holder authorizes a change.
