针对数据存储中心硬盘故障数据稀少造成的故障预测效果不佳的问题,面向自我检测分析与报告技术(self-monitoring analysis and reporting technology,SMART)数据信息的时序特征,提出一种通过数据增强解决不平衡问题的硬盘故障预测算法...针对数据存储中心硬盘故障数据稀少造成的故障预测效果不佳的问题,面向自我检测分析与报告技术(self-monitoring analysis and reporting technology,SMART)数据信息的时序特征,提出一种通过数据增强解决不平衡问题的硬盘故障预测算法。该算法利用长短期记忆网络改进传统的生成对抗网络,生成包含故障恶化趋势信息的序列段数据,解决了数据集不平衡问题。同时,为进一步提高预测性能,预测模型融合了时序注意力机制和特征注意力机制,挖掘不同SMART特征和时间步对硬盘故障恶化过程的敏感程度。此外,在特征选择阶段结合了多种典型特征选择算法来选取关键特征。在真实硬盘数据集上进行了实验验证,结果表明,所提算法的准确率、召回率和F 1值均有较大提升。展开更多
According to relevant data statistics, each year, nearly 70% of users have disc data loss because of misuse, viral damage, physical damage and hardware failure, bringing irreparable damage to enterprises, institutions...According to relevant data statistics, each year, nearly 70% of users have disc data loss because of misuse, viral damage, physical damage and hardware failure, bringing irreparable damage to enterprises, institutions and individuals. So data recovery technology has attracted wide attention of users and how to use data recovery technology to help to recover lost data and to minimize the loss has become an urgent need. This paper starts from the software and hardware failure to analyze two aspects of hard disk data loss, data storage structure theory, and combined with practical experience, it elaborates data damage type and related recovery methods.展开更多
Redundant array of inexpensive disk （RAID）10 is known as the most reliable disk array architecture to tolerate up to half of total disk failures, but failure of two disks in the same mirror set still results in data...Redundant array of inexpensive disk （RAID）10 is known as the most reliable disk array architecture to tolerate up to half of total disk failures, but failure of two disks in the same mirror set still results in data loss. In this paper, we propose a new disk array architecture, mirroring and parity protected RAID （MP-RAID）, which combines both mirroring and parity techniques to further improve reliability of disk arrays. The main idea behind MP-RAID is to protect the data by both mirroring and parity techniques： keep two copies of data in the same mirror set and update the parity block in the log disk within the same parity groups. Reliability analysis shows that the reliability of MP-RAID, in terms of mean time to data loss （MTTDL）, is much better than RAID10 and RAID5.展开更多
文摘According to relevant data statistics, each year, nearly 70% of users have disc data loss because of misuse, viral damage, physical damage and hardware failure, bringing irreparable damage to enterprises, institutions and individuals. So data recovery technology has attracted wide attention of users and how to use data recovery technology to help to recover lost data and to minimize the loss has become an urgent need. This paper starts from the software and hardware failure to analyze two aspects of hard disk data loss, data storage structure theory, and combined with practical experience, it elaborates data damage type and related recovery methods.
基金Project supported by the National Basic Research Program of China (Grant No.2004CB318201)the National High-Technology Research and Development Program of China (Grant No.2008AA01A401)the Changjiang Innovative Group of Ministry of Education of China (Grant No.IRT0725)
文摘Redundant array of inexpensive disk （RAID）10 is known as the most reliable disk array architecture to tolerate up to half of total disk failures, but failure of two disks in the same mirror set still results in data loss. In this paper, we propose a new disk array architecture, mirroring and parity protected RAID （MP-RAID）, which combines both mirroring and parity techniques to further improve reliability of disk arrays. The main idea behind MP-RAID is to protect the data by both mirroring and parity techniques： keep two copies of data in the same mirror set and update the parity block in the log disk within the same parity groups. Reliability analysis shows that the reliability of MP-RAID, in terms of mean time to data loss （MTTDL）, is much better than RAID10 and RAID5.
