BackBlaze Drive Data Redux

Node Operators
,
1

Since some posters have indicated that BackBlaze has relevant drive testing data for SNOs to reference, I have collected some relevant data regarding the drives which BackBlaze has reported data. Hopefully this information is useful for SNOs to reference when deciding on which drives to run and configuration of those drives. Please check my math and other numbers. The drive stats are pulled from the manufacturer’s drive datasheets.

The following table is an expanded table of found on BackBlaze’s web page. It does not contain the number of failed drives per grouping, I’ll add that column if people want me to…

Model           Market      Units  Price Details / Warranty / bit read failure rate
-------------------------------------------------------------------
HMS5C4040ALE640 Enterprise  2852   $172  SATA 6Gbps  3 year 10^14  24/7 180TB/year
HMS5C4040BLE640 Enterprise  12746  $130  SATA 6Gbps  3 year 10^14  24/7 180TB/year
HUH728080ALE600 Enterprise  1000   $258  SATA 6Gbps  5 year 10^15  24/7 
HUH721212ALE600 Enterprise  1560   $300  SATA 6Gbps  5 year 10^15  24/7
HUH721212ALN600 Enterprise  10859  $300  SATA 6Gbps  5 year 10^15  24/7
ST4000DM000     Desktop     19211  $78   SATA 6Gbps  1 year 10^14  unknown 55TB/year   only common Desktop Drive
ST6000DX000     Desktop     886    $235  SATA 6Gbps  1 year 10^14  unknown 55TB/year
ST8000DM002     Desktop     9809   $349  SATA 6Gbps  1 year 10^14  unknown 55TB/year
ST8000NM0055    Enterprise  14447  $210  SATA 6Gbps  5 year 10^15  24/7
ST10000NM0086   Enterprise  1200   $245  SATA 6Gbps  5 year 10^15  24/7
ST12000NM0007   Enterprise  37004  $308  SATA 6Gbps  5 year 10^15  24/7 outlier for enterprise (known problem)
ST12000NM0008   Enterprise  7215   $286  SATA 6Gbps  5 year 10^15  24/7
MD04ABA400V     Desktop     99       -   SATA 6Gbps  5 year 10^14  24/7
MG07ACA14TA     Enterprise  3619   $481  SATA 6Gbps  5 year 10^16  24/7 550 TB/year

Newegg prices listed.

Desktop Percent Failure = 1.789% (not annualized)

ST12000NM0007 failed controller, known problem

Enterprise drive failure without outlier = 375 / 55498 = 0.675% (not annualized)

Desktop Manufacturer Split

Toshiba = 99 = 0.329%
Seagate = 29906 = 99.67%

Failure rate of the most common Desktop drive = 2.0925%

Common Desktop Drive manufacturers not tested:

Western Digital, about 50% market share
Hitiachi
Dell
HP
Panasonic
Fujitsu

Zero NAS (mid-level spec) drives tested

Conclusion:

BackBlaze drive statistics do not show a random representative cross section of the Desktop hard drive market. Hard drives are not interchangeable devices. Each hard drive and hard drive manufacturer are very different. Test data for one drive can not be used to infer anything specific about any other drive. For all intents and purposes, there is only a single model of Desktop hard drive listed. So, if a given SNO has or is planning on using the Seagate ST4000DM000 4GB drive, that SNO may rely on the BackBlaze data set for expected drive failure and performance.

It should be noted that except for the one outlier Enterprise drive, the manufacturer datasheets showed a failure rate very similiar to the failure rate found with the tested drives. It should also be noted that the most common desktop drive had the highest non-annualized failure rate of any drive except for the Enterprise with the known manufacturer defect. Furthermore, the drive with the highest failure rate had the shortest warranty period of just one year.

5 Likes
2

I love this data! Thanks for collecting that.

Could you elaborate a little on what you mean by the failure rate not being annualized? That seems to me that drives being used longer would return worse results because they’ve had more time to fail. But perhaps I’m misunderstanding.

For reference, these are the life time annualized results per model as backblaze reported them.

image
image782×664 84.7 KB

It kind of sucks that they don’t include WD as well. I’m sure they have their reasons. I would also have loved some NAS class models in there. But the spread between enterprise and desktop class isn’t that big and I’m sure NAS is somewhere in between. I have no reason to believe WD would be significantly worse than the HDD’s on this list, but that is mostly based on anecdotal evidence. Generally WD has a very good reputation, especially the WD Red line which seems to still be the go to advise for a lot of people when suggesting HDD’s for a NAS. I’m not sure if that’s because it was just the first prominent line to specialize for NAS or something else as well, but I have very good experiences with Seagate Ironwolfs as well.

With only 3 HDD manufacturers left, the ones that remain will have to be competitive or they will disappear as well. Sure there are differences between HDD models and manufacturers. But I doubt there is a single modern HDD that would get an AFR over 3%. Even when looking at the backblaze results the ST4000DM000 seems to be an outlier in the desktop HDD category. All other models seem to perform significantly better.

So if anyone knows of another good source of actual failure rates, please link us to this information as well so we can see if we can close some of the gaps in this data.

3

Backblaze stated their reason for not using Western digital, others is they couldnt get big enough bulk purchases of the drive from the manufacturer

1 Like
4

I use WD long time, have no wories with them. Most WD perple.