Dataset · enterprise SSD replacements · FAST 2020
Enterprise SSD replacements ran at 0.22% a year across 1.4 million NetApp drives, with firmware the biggest swing
Maneas et al., FAST 2020, mined NetApp support messages for about 1.4 million SSDs (3 manufacturers, 18 models, January 2017 to May 2019). The average annual replacement rate was 0.22%, from 0.07% to nearly 1.2% by model. Early firmware versions of some families were replaced 2 to more than 10 times as often as later ones, and the chance of another replacement within a week in a RAID group was 9.39% against 0.0504% in a random week. Cross-check rows come from Google (FAST 2016).
Download CSV17 rows · 7 columns · CSVMethod
The CSV transcribes figures printed in Maneas et al. (FAST 2020) and two cross-check rows from Google (FAST 2016). It does not read chart values. The annual replacement rate is replacements divided by device years. A replacement is a drive marked as failed and swapped out, often while it still works.
Limits
- Replacement is an operator action and about a third of replacements are predictive, so the annual replacement rate is not a data-loss rate.
- The reason is missing for 40% of replacement events and the paper assumes the rest are spread in proportion.
- The firmware effect is within one family; the cause is an explanation in the paper, not a measurement.
- Rated life used is a truncated integer, so many drives show 0.
- The Google rows measure the share of drives removed in 4 years, not an annual rate, and cover another fleet.
Table
| Measure | Value low | Value high | Unit | Scope | Note | Citation |
|---|---|---|---|---|---|---|
| average annual replacement rate | not stated | 0.22 | percent per year | All studied NetApp SSDs | Replacements divided by device years. | maneas-fast20-pdf |
| annual replacement rate across models | 0.07 | 1.2 | percent per year | 18 models | Printed as nearly 1.2 at the top. | maneas-fast20-pdf |
| annual replacement rate of II-G 15TB | not stated | 0.53 | percent per year | One model | Similar specifications to II-C 15.3TB. | maneas-fast20-pdf |
| annual replacement rate of II-C 15.3TB | not stated | 1.13 | percent per year | One model | One of two outlier models excluded from some figures. | maneas-fast20-pdf |
| replacements caused by SCSI errors | not stated | 32.78 | percent of replacements | Reason type normalized for the 40% of events with no reason | Printed as about 33%. | maneas-fast20-pdf |
| replacements where the drive became unresponsive | not stated | 0.60 | percent of replacements | Same normalization | maneas-fast20-pdf | |
| length of the infant-mortality period | 20 | 40 | percent of a 5-year life | eMLC and 3D-TLC drives | Failure rates rose for 12 to 15 months and then fell for 6 to 12 months. | maneas-fast20-pdf |
| higher replacement rate with less than 1% rated life used | not stated | 1.25 | times | eMLC drives | Compared with drives above 1% of rated life used. | maneas-fast20-pdf |
| ARR drop from FV2 to FV3 in family II-A | not stated | 8 | times | Firmware versions within one family | Printed as factor 8X. | maneas-fast20-pdf |
| ARR drop from FV2 to FV3 in family II-F | 10 | times | Firmware versions within one family | Printed as more than 10X so 10 is a lower bound. | maneas-fast20-pdf | |
| ARR drop from FV1 to FV2 in family I-B | 2 | times | Firmware versions within one family | Printed as more than 2X so 2 is a lower bound. | maneas-fast20-pdf | |
| drives with an empty defect list | not stated | 99.04 | percent of drives | Defect list (g-list) of unrecoverable-error blocks | Drives with a non-empty list were replaced more often. | maneas-fast20-pdf |
| RAID group replacement in a random week | not stated | 0.0504 | percent per week | All RAID groups | maneas-fast20-pdf | |
| another replacement within a week of a replacement | not stated | 9.39 | percent per week | Same RAID group | More than 180 times the random-week figure. | maneas-fast20-pdf |
| consecutive replacements within one week | not stated | 52 | percent of consecutive replacements | RAID groups with more than one replacement | 46% were within one day. | maneas-fast20-pdf |
| Google flash drives permanently removed within 4 years | 5 | 10 | percent of drives | Google data centers 10 models | About 5 for most models and about 10 for the worst. Not an annual rate. | schroeder-fast16-pdf |
| Google flash drives with uncorrectable errors in four years | 20 | percent of drives | Same population | Printed as more than 20. | schroeder-fast16-pdf |
Columns
- measure
- The quantity as the source names it.
- value_low
- Lower end of a printed range. Empty when the source prints a single value.
- value_high
- Single printed value, or the upper end of a range. For 'more than half' it is the stated bound.
- unit
- Unit of the value columns: disks, blocks, files, percent of disks, percent of mismatches, or probability.
- scope
- Population and window the value applies to.
- note
- What the value is not, or the source wording behind it.
- citation_id
- Id of the opened source in the citations list.
License
Small derived table of figures printed in Maneas et al. (FAST 2020) and Schroeder et al. (FAST 2016), with credit. Not a Creative Commons license. The USENIX proceedings state that rights to individual papers remain with the author or the author's employer. This file is not a copy of the papers and not the underlying telemetry.