All mechanical hard drives fail after a while. For consumers, it may not be that big of an issue as long as data is properly backed up, but for businesses, a higher failure rate may lead to extra maintenance costs and downtime.
As a cloud storage company, Backblaze has lots of drive, and by a lot, we mean over 140,000 hard drives from different vendors, and they happen to provide quarterly and yearly updates to the failure rates of different drives. The latest one is for Q2 2020.
Drive days represent the number of days, hard drives were operational during the period, basically drive count x (365/4) – maintenance/replacement days. AFR stands for “Annualized Failure Rate”. HGST (Hitachi Global Storage Technologies) was a manufacturer of hard drive acquired by Western Digital in 2012, but the company still sells HGST parts… More on that later.
There were three drive models without any failure during the April-June 2020 quarter: HGST HUH728080ALE600 (8TB), Seagate ST6000DX000 (6TB), and Toshiba MD04ABA400V (4TB). BackBlaze points out the latter did not have a single failure since Q4 2018 or 54,054 drive days.
The good news is that the AFR is constantly decreasing dropping from 1.8% in Q2 2019 to under 1% for the first time in Q2 2020. What we don’t see in this table is the average life of the drives, so it may be important to have a look at a longer-term chart.
If you care about reliability HGST hard drives seem to be the way to go. There’s a lot of variation for some other brands due in part because the company has less WDC and Toshiba drives. Wait… but why does Western Digital Corporation (WDC) chart stop in Q1 2019?
That’s because Backblaze simply stopped using drives from Western Digital with the official reason being WD could not meet price and/or availability requirements from the company. Seagate had over double the failure rate of HGST drives, but for some reason, they get on adding more over the years. Seagate AFR still dropped to 0.9% so it’s improved a lot.
The cumulative annualized hard drive failure chart smoothens the results and does not look good at all for Western Digital, with HGST and Toshiba drives being the most reliable based on Backblaze data. But as Andy Klein, Director of Compliance at Backblaze, explains it’s important to focus on the drive model instead of just the manufacturer. You’ll find all historical data on Backblaze website.
Via Liliputing
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
It would be interesting to see how different these charts would look if drives with under 200k hours were excluded.
Note that this does not say much about the quality. In the original report they also report 5 failures on the Toshibas between April 2013 and jun2020. (with 252.766 drive days)
Typically the failure pattern for hard disk is that some fail very early in their lifetime, then the failure rate drops and at some point the failure rate starts to increase. (kind of a bath-tub curve).
It may well be that they added these 99 drives all at roughly the same time (getting a batch of 100?) and that they are now at the floor of the bath-tub curve.
If drives are added over a longer period of time there will be more failures due to the initial fail rate.
That does not say the toshiba’s are bad. The data is just not detailed enough to draw a firm conclusion (and anyway the number of drives (99) on the small side to draw conclusions)
PS: nice article and interesting findings
Their quarterly data presentation is mostly an example of ‘statistics gone wrong’ targeted at consumers. If you want to derive information from data you need to filter out stuff without statistical significance and need to weigh specific data. If they would do this the graphs would look differently. Differently in a way no consumer would want to look at the graphs any more.
But hey, who cares? All that matters are some graphs and percentages people are talking about so kudos to the Backblaze marketing team since they get free press coverage world-wide every 3 months.
For people wanting to rely on their reports think about your use case first:
Ah yes
There are lies, damned lies and statistics 😉
And actually the most stressful part of a disk is the poweron behavior. That is not something that is found from data like this, but it is important for your desktop system.
And if your drive is not used for a prolonged time it may not even spin up any more (maybe due to the bearings giving more friction)
Are the Seagate and Toshiba drives as loud as the HGST?
HGST seem to be notoriously loud for typical home NAS use
Your question shows why these Backblaze ‘statistics’ are so terribly misleading. HGST is not a HDD manufacturer but since more than 8 years just a Western Digital brand being replaced by the WD brand starting two years ago.
As such HGST was just a WDC label for enterprise drives that usually are louder than desktop/consumer or homeuser NAS drives.