Study of Failure Hazard Degree in Large Data Centers

Author Name(s): Iskandar N. Nasyrov, Ildar I. Nasyrov, Rustam I. Nasyrov, Bulat A. Khairullin
Author Email: ecoseti@yandex.ru

Abstract

The change in SMART values that occurred during operation, namely, parameters characterizing the reliability of hard magnetic disks in computers, is considered. It is precisely those parameters that are considered critical due to the fact that if their values increase, the probability of failure of information storage devices on hard drives increases. The scientific task of the study is to establish the priority of the specified parameters in failed hard drives from the standpoint of assessing the reliability of information storage devices of various manufacturers. The study analyzed the information storage devices of the brands HGST, Hitachi, Samsung, ST, Toshiba, WDC, operated in the largest commercial data center Backblaze. The analysis revealed the following priority of critical parameters: 5 Reallocated sectors count (number of reassigned sectors), 196 Reallocation event count (number of successful and unsuccessful attempts at the reassignment), 197 Current pending sector count (number of sectors, which are been candidates for the replacement), 1 Read error rate (frequency of errors during reading of data from the disk), 9 Power-on hours (the number of hours, carried out in the switched-on state), 7 Seek error rate (frequency of errors during positioning of the unit of magnetic heads), 10 Spin-up retry count (number of repeated attempts at the spin up of disks to the operating speed if the first attempt was unsuccessful). It is shown that the very existence of the values of the parameters considered depends on the manufacturer of the hard drives. The ranking of information storage devices by the degree of reliability of implementation using the priority parameters identified as a result of the study is proposed

Introduction

To ensure the security of information, it is necessary to copy the data from the unreliable to a new and reliable drive in a timely and complete manner. For this purpose, usually SMART technology is used (self-monitoring, analysis and reporting technology [1]) for internal assessment of the state of the hard disk of the computer, as well as a way to predict the possible failure of it. The paper considers the change occurred due to the operating time of the values of SMART-parameters characterizing the reliability of hard magnetic disks in computers. The parameters selected are critical in the sense that if their values increase, the probability of failure of the hard disk drives increases. The scientific objective of the study is to establish the priority of these parameters in failed hard drives from the standpoint of assessing the reliability of information storage devices of various manufacturers. The study analyzed drives of brands HGST, Hitachi, Samsung, ST, Toshiba, WDC, operated in the largest commercial data center Backblaze. The analysis revealed the following priority of critical parameters (in descending order): 5 Reallocated sectors count (number of reassigned sectors),196 Reallocation event count (number of successful and unsuccessful attempts at the reassignment), 197 Current pending sector count (number of sectors, which are been candidates for the replacement), 1 Read error rate (frequency of errors during reading of data from the disk), 9 Poweron hours (the number of hours, carried out in the switched-on state), 7 Seek error rate (frequency of errors during positioning of the unit of magnetic heads), 10 Spin-up retry count (number of repeated attempts at the spin up of disks to the operating speed if the first attempt was unsuccessful). It is shown that the very presence of the values of the parameters examined depends on the manufacturer of hard disks. The ranking of information storages according to the degree of reliability to carry out with the application of that revealed as a result of a study of parameters priority is proposed.

Conclusion

Another argument in support of the use of these parameters is that to characterize the state of the memory space, the same parameters are used with the same meaning in solid-state storage devices [9, 10]. Naturally, the positioning parameters of the write/read heads or disk rotation mechanisms in solid-state drives are not used, although the numbering remains the same. Therefore, in the future, when large data centers move to store information in drives of the indicated type, no additional changes other than the exclusion of parameters 7 and 10 will have to be made to the developed algorithm. Backblaze experts report the results of their research at conferences [11, 12], however, topics of algorithms for ranking information storage devices by the degree of failure hazard do not affect. Therefore, the developed algorithm has scientific novelty and allows the individual assessment of the reliability of information storage devices to be solved using the identified priority parameters

793 total views, no views today

Download PDF File

About the author: admin