Reliability Test For Enterprise Hard Drives

This is a project to estimate reliability of enterprise HDD/SSD drives by the analysis of SMART reports from servers. The primary aim of the project is to find drives with longest power-on hours (POH) and minimal number of errors, i.e. maximal MTBF (mean time between failures).

Everyone can contribute to this report by installing hw-probe to their OpenVZ 7 or Virtuozzo 7 servers:

sudo -E hw-probe -all -upload

Total drives: 42893.

Contents

1. About
2. HDD by Model
3. HDD by Family
4. HDD by Vendor
5. SSD by Model
6. SSD by Family
7. SSD by Vendor
8. NVMe by Model
9. NVMe by Vendor

About

The structure of the reports directory is the following:

{KIND OF DRIVE}/{VENDOR}/{MODEL PREFIX}/{MODEL}/{DRIVE ID}

( e.g. HDD/Seagate/ST9500/ST9500620NS/B0744A84D66D )

Based on SMART data we determine most reliable drive models and vendors by the following formula:

MTBF = MAX( Power_On_Hours / ( 1 + Number_Of_Important_Errors ) ) / ( 365*24 )

( i.e. MTBF = "Years before/between errors" )

Number_Of_Important_Errors — number of important errors that can indicate a drive failure:

• Current_Pending_Sector
• ECC_Uncorr_Error_Count
• End-to-End_Error
• Offline_Uncorrectable
• Reallocated_Event_Count
• Reallocated_Sector_Ct
• Reported_Uncorrect
• Soft_Read_Error_Rate
• Spin_Retry_Count
• Total_Pending_Sectors
• Unc_Soft_Read_Err_Rate

One can estimate reliability of a hard drive (i.e. probability that the unit will work for some time T without failure) by:

Reliability(T) = exp(-T/MTBF)

The list of important SMART attributes is taken from recent studies of Google [1], Backblaze [2] and Acronis [3]. If an attribute exceeds the value of 1000 then it's counted as '1000 + log10(X - 999)'.

The list of tracked attributes and MTBF calculation method can be discussed. Please suggest your ideas in "issues".

You can perform your own analysis of collected reports if needed.

• [1] Google, "Failure Trends in a Large Disk Drive Population" (https://research.google.com/archive/disk_failures.pdf)
• [2] Backblaze, "Hard Drive SMART Stats" (https://www.backblaze.com/blog/hard-drive-smart-stats/)
• [3] Acronis, "Acronis Drive Monitor: Disk Health Calculation" (https://kb.acronis.com/content/9264)

HDD by Model

Please take all columns into account when reading the table. Pay attention on the number of tested samples and power-on days. Simultaneous high values of both MTBF and errors are possible if only rare drives in the subset encounter errors.

Days — avg. days per sample, Err — avg. errors per sample, MTBF — avg. MTBF in years per sample.

See complete list of tested HDD samples in the Appendix 1 (All_HDD.md).

HDD by Family

Please take all columns into account when reading the table. Pay attention on the number of tested samples and power-on days. Simultaneous high values of both MTBF and errors are possible if only rare drives in the subset encounter errors.

Days — avg. days per sample, Err — avg. errors per sample, MTBF — avg. MTBF in years per sample.

HDD by Vendor

Please take all columns into account when reading the table. Pay attention on the number of tested samples and power-on days. Simultaneous high values of both MTBF and errors are possible if only rare drives in the subset encounter errors.

Days — avg. days per sample, Err — avg. errors per sample, MTBF — avg. MTBF in years per sample.

SSD by Model

Please take all columns into account when reading the table. Pay attention on the number of tested samples and power-on days. Simultaneous high values of both MTBF and errors are possible if only rare drives in the subset encounter errors.

Days — avg. days per sample, Err — avg. errors per sample, MTBF — avg. MTBF in years per sample.

See complete list of tested SSD samples in the Appendix 2 (All_SSD.md).