For the second year running, the cloud storage firm Backblaze has published a report detailing the reliability of the over 41,000 drives in their facility. Backblaze is noted for using consumer-grade drives in its data centers, as opposed to the more costly enterprise-grade drives, which the company contends are not worth the extra cost.
Statistics were released for 17 drive models (the report doesn’t include drive models that Backblaze has less than 45 of). The results of those statistics, however, lead to a rather startling finding: Some Seagate drives have a higher failure rate than drives from other manufacturers, with one model having a failure rate over 40%.
Failure by numbers: What doesn’t work?
For failure frequency, the top three drives are the 3 TB Seagate Barracuda 7200.14 at 43.1%, the 1.5 TB Seagate Barracuda 7200.11 at 23.5%, and the 1.5 TB Seagate Barracuda LP at 9.5%. It’s important to note that the average age of those drives is 2.2 years, 4.7 years, and 4.9 years, respectively. Of those three, the Barracuda LP — which runs at 5,900 RPM — is more reliable than the 7,200 RPM drives with failure rates in the double digits. This isn’t to say all Seagate drives are faulty, as its 4 TB Desktop HDD — also running at 5,900 RPM — has only a 2.6% failure rate across over 12,000 drives at an average age of 0.9 years.
But the big picture is rather troubling: Slower drives being more reliable than desktop drives, and the drives altogether having a nontrivial failure rate that is several times that of other drives. In a statement to The Register, a representative from Seagate noted:
It appears that Backblaze is reporting data from the same sample of drives from last year, which continues to be inconsistent with data received from other customers, and our large OEM installed base. …Yet, as with previous data reported by Backblaze, desktop-class drives and some external drives were purchased and used in enterprise-class workloads – which they were NOT designed for nor tested to support. Therefore, we agree with Backblaze’s comment that “It may be that those drives are less well-suited to the data center environment. Or it could be that getting them by drive farming and removing them from external USB enclosures caused problems.”
Does either side have a valid argument?
Seagate’s argument has one particular issue that is difficult to ignore: The comparison made by the Backblaze report does not compare enterprise drives to consumer drives; the failure rates and comparisons being made are only between consumer drives of major vendors. The more important factor is that of the Seagate drives represented, the drives failing more frequently are last-generation drives with smaller capacities.
In my opinion, Seagate’s argument does not withstand scrutiny particularly well, as the differences between enterprise and consumer drives are primarily in the warranty — the physical process through which data is recorded on the drive platters is functionally identical on a mechanical level between different SKUs of the same generation drive. This is not a rejection of a particular storage method, such as Seagate’s upcoming HAMR technology.
Of course, firmware differences do exist — drives that have aggressive timing for head parking can have an artificially shorter life when used in RAID arrays, or other high-access use cases. This issue is far more pronounced with drives marketed as “Eco” or “Green” drives, which typically have a slower speed of 5,400 or 5,900 RPM. For faster desktop drives, the differences between the two classes of products are primarily in the warranty — consumer drives typically have two-year warranties, while enterprise drives have warranties of three to five years.
Even so, the failures of those drives have not dissuaded Backblaze from buying current-generation Seagate products, and the final verdict from its report states that “if you are looking for good drive at a good value, it’s hard to beat the current crop of 4 TB drives from HGST and Seagate.” As previously mentioned, Backblaze has deployed over 12,000 of Seagate’s new 4 TB drives, and are testing one rack of 45 6 TB drives, which have been running for a few months without failure. In the report, Backblaze notes that Seagate’s 4 TB drives have a 2.6% failure rate in the first year, while the 3TB Barracuda 7200.14 had a 9.3% failure rate.
What’s the real takeaway from this report?
Owing in large part to their mechanical nature, all drives will eventually fail. Having adequate off-site backups is the best method to protect against that failure. Bearing that in mind, choosing a drive that is affordable — though not necessarily the cheapest available — and has adequate reliability is a balancing act for your budget and peace of mind.
For people interested in studying the data used in this report, Backblaze has freely released the raw data detailing the active hard drives in its data center for 2013 and 2014 in CSV format, with instructions for converting it to SQLite.
What’s your view?
Do you have a vendor or drive model that you prefer, or one that you avoid at all costs? For desktop users, is there a standalone RAID solution you prefer to use to protect against individual drive failure? Tell us about your disk deployment and thoughts in the comments.