SSD and HDD Failure Detection using Advance Deep Learning Algorithms

Muralidhar, Sandesh (2022) SSD and HDD Failure Detection using Advance Deep Learning Algorithms. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (757kB) | Preview

Abstract

Data centers are under demand to provide ever-more-efficient services due to the growing need for data processing and storage. But these services effectiveness may be effected by their dependence on hard drives—in particular, solid-state and magnetic drives, which are now among the most widely used types of data storage. Occasionally, these devices may fail and cause permanent data loss, which would violate contractual service level commitments and cause financial harm to both the customer and the hosting provider. This research aims to increase the cloud storage service quality by predicting SSD (Solid-State Drive) and HDD(Hard disk drives) failures using machine learning methods. Hard disks failure is a significant cause of application failures which leads to potential data loss and downtime. This research explores the application of deep learning methods and ensemble algorithms for detecting Hard drives failure prediction. Blackblaze dataset which has details of SMART parameters has been used for Research. This Research also explores on finding top parameters effecting the drive failure. It explores deep learning techniques, especially Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and a hybrid Conv-GRU model. The study takes into account both the spatial and temporal facets of the dataset to resolve the urgent demand for trustworthy forecasting algorithms as a solution. To assure effectiveness, the research employs a complete methodology that begins with thorough feature engineering and includes the extraction of the most important characteristics. A detailed correlation analysis of SMART parameters has also been performed in this research. Three deep learning models —Convolutional Neural Network (CNN), Gated Recurrent Unit(GRU), and a hybrid CNN-GRU—are applied to study it’s efficiency in prediction of failures. The Conv-GRU model is the best-performing algorithm of the ones that were looked into; it performs better in terms of accuracy, precision, recall, and F1-score metrics. Its superior performance over CNN and GRU equivalents is largely due to its capacity to combine spatial and temporal information. The outcomes of this research would contribute to improve efficient storage management, hence the overall availability of datacenters and quality of cloud storage service.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Arun, Shreyas Setlur UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > QA Mathematics > Algebra > Algorithms > Computer algorithms Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	26 Mar 2025 15:58
Last Modified:	26 Mar 2025 15:58
URI:	https://norma.ncirl.ie/id/eprint/7338

Actions (login required)

View Item