Dataset · SSD failures · SYSTOR 2016

38% of failed datacenter SSDs showed none of the four SMART symptoms; the ones that did failed 3 to 20 times as often

Narayanan et al., SYSTOR 2016, studied over half a million SSDs in one operator's datacenters over nearly 3 years. 62% of failed devices had shown at least one of four SMART symptoms and 38% none; devices with a symptom failed 3 to 20 times as often. A classifier reached 87% precision and 71% recall with 5-fold cross-validation. Cross-check rows come from NetApp (FAST 2020) and Backblaze (2016).

Download CSV14 rows · 7 columns · CSV

File /data/datasets/ssd-failures-datacenter-symptoms.csv · JSON metadata · All datasets · Human-readable note on hesela.com

Method

The CSV transcribes figures printed in Narayanan et al. (SYSTOR 2016) and two cross-check rows from NetApp (FAST 2020) and Backblaze (2016). It does not read chart values. AFR is the share of devices with failures divided by device years. A failure here is a fail-stop that takes a server down.

Limits

One operator's fleet and workloads; model names are anonymized.
Failure means a fail-stop that takes the server down; about 80% ended in replacement.
Per-model AFR values are in charts, so only printed text figures are used.
The classifier was scored with random 5-fold cross-validation, not a split by time.
The cross-check rows are other fleets with other definitions.

Table

14 rows, the same rows as the CSV. Value low and value high are the printed range ends, or a single printed value in the high column. Units are in the unit column. Sources are the pages opened on 2026-10-10.
Measure	Value low	Value high	Unit	Scope	Note	Citation
SSDs studied	500000		devices	5 model groups and one operator	Printed as over half a million.	narayanan-systor16-pdf
vendor-quoted annual failure rate for consumer models	0.61	0.73	percent per year	Specification range shown in the AFR chart		narayanan-systor16-pdf
observed AFR above specification for some models	not stated	70	percent above the quoted rate	Models 1-B and 1-C exceeded the range	Printed as as much as 70.	narayanan-systor16-pdf
replacement after an SSD failure ticket	not stated	79	percent of tickets	Datacenter tickets in the study	Hard disk tickets were 11%.	narayanan-systor16-pdf
AFR increase when a symptom is present	3	20	times	Four SMART symptom categories	Data errors up to 20 times.	narayanan-systor16-pdf
AFR increase with reallocations or SATA downshift	not stated	4	times	Compared with devices without the symptom		narayanan-systor16-pdf
AFR increase with program or erase failures	not stated	2.75	times	Compared with devices without the symptom		narayanan-systor16-pdf
failed devices with at least one of the four symptoms	not stated	62	percent of failed devices	Fail-stop failures	Printed as around 62. The other 38% showed none.	narayanan-systor16-pdf
healthy devices with data errors	not stated	1	percent of healthy devices	Same fleet	About 12% of failed devices had them.	narayanan-systor16-pdf
AFR increase with average writes per day	2	4	times	Older models 1-A and 1-B and 1-C	Not seen for model 1-D.	narayanan-systor16-pdf
classifier precision	not stated	87	percent	Failed versus healthy devices	5-fold cross-validation with SMOTE over-sampling.	narayanan-systor16-pdf
classifier recall	not stated	71	percent	Same classifier		narayanan-systor16-pdf
average annual replacement rate of enterprise SSDs	not stated	0.22	percent per year	NetApp and about 1.4 million SSDs	Models ranged from 0.07 to nearly 1.2.	maneas-fast20-pdf
Backblaze failed drives with one or more of five SMART counts above zero	not stated	76.7	percent of failed drives	67814 drives in 2016	23.3% showed none.	backblaze-smart-2016

Columns

measure (string): The quantity as the source names it.
value_low (number): Lower end of a printed range. Empty when the source prints a single value.
value_high (number): Single printed value, or the upper end of a range. For 'more than half' it is the stated bound.
unit (string): Unit of the value columns: disks, blocks, files, percent of disks, percent of mismatches, or probability.
scope (string): Population and window the value applies to.
note (string): What the value is not, or the source wording behind it.
citation_id (string): Id of the opened source in the citations list.

License

Small derived table of figures printed in Narayanan et al. (SYSTOR 2016), Maneas et al. (FAST 2020), and a Backblaze post of 6 October 2016, with credit. Not a Creative Commons license. The ACM and USENIX papers remain with their authors or publishers. This file is not a copy of the papers and not the data.

Sources

SSD Failures in Datacenters: What? When? and Why?, Narayanan et al., SYSTOR 2016 (author PDF, ACM article 7, pages 1 to 11) (accessed 2026-10-10)
A Study of SSD Reliability in Large Scale Enterprise Storage Deployments, Maneas, Mahdaviani, Emami, and Schroeder, FAST 2020 (USENIX PDF, pages 137 to 149) (accessed 2026-10-10)
What SMART Stats Tell Us About Hard Drives, Backblaze, 6 October 2016 (accessed 2026-10-10)