Hard drive reliability

[SlayVus]

So I was looking at the stats of WD's line up of all their 3.5" hard drives.

This is the information that I found from their site for their major line ups.

Non-recoverable read errors per bits read
Black, Blue, Red - < 1 in 10^14
Purple - < 10 in 10^14
Red Pro and Se - < 10 in 10^15
WD Re - < 10 in 10^16

I understand that 10^15 is 10 times more reliable than 10^14, but how reliable is <10 in 10^14 compared to <1 in 10^14? Or even <1 in 10^14 to <10 in 10^15?

 

[mtrupi]

I'm not a HDD expert but I found one place that attempted to put this in real terms. I didn't verify everything there but it looked reasonable to me.
http://superuser.com/questions/5169...-unrecoverable-read-error-during-raid-rebuild

 

[drescherjm]

how reliable is <10 in 10^14 compared to <1 in 10^14?

(<1 in 10^14) is 10 times (<10 in 10^14)

Or even <1 in 10^14 to <10 in 10^15?

The same.

Also when we are talking about this in the real world you do not see a uniform 1 bit error per 10^14 on drive that is operating properly. At least I do not ever see this in my weekly scrubs of >70TB of data. Its more of drives that are working have a much lower error rate and bad drives have a much higher error rate. This 1 bit error per 10^14 is an average over the whole population of good and failing drives of a particular model.

 

[SlayVus]

Wait? How can <10 in 10^14 be ten times more reliable, but <10 in 10^15 is the same as <1 in 10^14 when 10^15 is already inherently ten times more reliable?

 

[drescherjm]

I think I fixed that..

 

[devman]

Non-recoverable read errors per bits read
Black, Blue, Red - < 1 in 10^14
Purple - < 10 in 10^14
Red Pro and Se - < 10 in 10^15
WD Re - < 10 in 10^16

rewritten in non marketing speak

Purple - < 1 in 10^13
Black, Blue, Red, Red Pro and Se - < 1 in 10^14
WD Re - < 1 in 10^15