[NewsPoster]

Backblaze has released its latest data on hard drive failure rates, finding that the results for the last 12 months were the lowest it has recorded, at just 1.84 percent. Its data center currently utilizes drives from HGST, Seagate, Toshiba, and Western Digital, for a total of 61,523 drives, up from 56,224 at the end of 2015. The data backup specialists have now had their drives operational for a combined one billion hours, providing a wealth of data on consumer-level hard drive performance and reliability.



As it has previously found, hard drives manufactured by HGST are significantly more reliable than the competition, although reliability gains have been made across the board. The Backblaze data for the March quarter this year shows that Seagate drives have made the largest gains in reliability -- although this will need to be reviewed over the course of the full year, to see if it is in fact part of a significant trend showing improved performance and reliability.

Historically, Seagate's drives have typically failed at a rate of around 10 percent on average over the course of 12 months. Backblaze, however, uses a particularly high number of drives from Seagate compared to those from Toshiba and Western Digital, which could have skewed its failure numbers upwards -- but not the percentages.

HGST is once again the clear outlier for overall performance reliability, however. Even with a relatively large sample size of 22,731 drives, its failure rate is an impressively-low 1.03 percent. This is in line with its performance over previous years, and with roughly similar sample sizes. Toshiba comes in second for overall reliability, with a failure rate of 3.06 percent, although its sample size is limited to just 238 drives.

Coming in at third place is Western Digital, with failure rates running relatively high at 6.55 percent for the quarter, from a relatively small sample size of 1,691 drives. Again, this is roughly in line with its historical performance, which is running at around an average failure rate of eight percent.

Previous reports by Backblaze have been criticized for "abnormal" temperature and vibrational stresses put drives in its data center. However, compared to most environments where these drives would also typically operate, such as in everyday PCs and uncontrolled home environments, Backblaze's data center is much more consistently monitored. Further, as the drives it is reporting on all operate in the same conditions in its data center, each drive is subject to the same conditions -- lending validity to its data, relative sample sizes notwithstanding.

[GaryDeezy]

The sample size for Toshiba was too low to be considered, in my opinion.

[Makosuke]

I kind of want to sign up with Backblaze just in thanks for publishing this data for the world. After so many years of anecdotal reports, vocal opinions based on a single past experience, and generally useless "data", It's really refreshing to have a chunk of actual, relatively rigorous, real-world data with a statistically significant sample size to refer to when discussing drive reliability.

There are a few things not addressed, like if the particular (presumably relatively high vibration) environment might cause higher failure rates than the same drive in consumer use, and how sensitive any particular model is to being cycled on and off, but it's still vastly more valuable information than available from any other source.

It's also worth noting that the table tells you a lot more than that bar chart since there appears to be a fair amount of variability between models even from the same manufacturer; if you only look at drives with at least a thousand examples, there's a variability between 0% and 1% just between various 4TB models from HGST.