Do you spread hard drive purchases amongst different sellers?

[solarisguy]

I love you guys :D

 

[danb35]

It's a statement on how long a large group will last before a single drive failure occurs.

So the number's effectively meaningless, then. Without knowing how many of your particular model were tested, it tells you nothing at all. Even knowing that number, it tells you that n disks survived their first (MTBF/n) hours.

And your remark about warranty tells me the same thing--the MTBF spec is meaningless.

 

[solarisguy]

The parameters shown are some measurements and estimates that only attempt to visualize how a failure curve might look like.

They are describing populations, and somewhat similarly to lottery numbers, there is nothing whatsoever guaranteed about performance of an individual drive.

Again, the original question was about insurance :D . So I prefer to have a drive from a population that averages to have better survival estimates. But one has to learn that her or his exemplars of nominally superior drives could turn out to be worse reliability performers, than the inferior (and cheap) drives someone else has bought.

 

[danb35]

Well, no, the original question wasn't about insurance. It was about whether taking one specific step would decrease the chances of suffering a disk failure. Though I don't think hard numbers can really be obtained, it seems the consensus on this thread is that it likely wouldn't.

You have turned the question into whether taking a different step (to wit, selecting a different drive model) would significantly decrease the chances of suffering a disk failure. It's a fair question, but it's not what @Dat Sysadmin asked. And I remain unconvinced that the reds are significantly better than the greens, but I'll concede that WD thinks either that they are better, or that the price delta is enough to cover the warranty replacements.

As to MTBF, or more precisely MTTF... The term is fairly simple, and should be understandable with the ordinary definitions of the four words that comprise it. Yes, I understand that a proper (i.e., meaningful) average covers a large population, and it can't be used to predict the behavior of a single member of that population. But still:

• Mean: the common "average". The (arithmetic) mean of a set of values is the sum of those values, divided by the number of values.
• Time: Since I don't have a TARDIS, I consider time as linear, rather than as a big ball of wibbly-wobbly, timey-wimey stuff.
• Failure: I'd consider this a complete failure to operate, though less-catastrophic events would probably qualify as well.

So, on its face, the term MTTF means that if I have a sufficiently-large population of widgets, and I run them to failure, the average time that they lasted before failing should approximate the MTTF (the larger the population, the more closely it should approximate the MTTF). Any given widget in that population may fail immediately after being put into service, shortly thereafter, or may exceed the MTTF spec by a significant amount, but the average should be close.

In the real world, we all know this isn't true. If I buy 1000 drives, and run them 24/7/365, they will all fail long before 57 years. WD knows it too, or they'd offer a longer warranty than 2 years (there are any number of reasons that it would still be less than 57 years). If the common practice in the industry is to run a group of n widgets until one fails at h hours (or run h hours with no failures), and then specify the MTBF/MTTF as n x h hours, then (1) they're being intentionally deceptive in the use of the term, and (2) that spec is almost completely meaningless in projecting lifespan (i.e., reliability) of a product.

 

[solarisguy]

I am assuming that Volkswagen managers and managers in the hard drive manufacturing sector are different :D . If only due to existence of powerful companies (like for example Google) that purchase hundreds of thousands of their drives and any significant deviations from published specifications are bound to be discovered.

 

[Deleted47050]

I get what @solarisguy is saying and again, so far I have only used red drives exclusively, because ON PAPER they should be better for 24x7 operations However, I have seen many examples on the forum of people using green drives successfully which, to be honest, made me question if it was really worth spending 30% more for the Red drives.

Still, this is not really the point :P my original question would apply to a batch of red drives as well. Apparently, spreading purchases looks overrated, which is what I wanted to check.

 

[Robert Trevellyan]

made me question if it was really worth spending 30% more for the Red drives

When I think about how long I expect my system to last, and amortize the cost difference of the drives over that period, I question whether it's really worth saving the % difference for a drive that isn't intended for NAS use.

Also, TLER.

spreading purchases looks overrated

I spread mine, by purchasing each time the price hit a low, but I was in no hurry to get my current system built.

Owners of dead and dying 3TB Seagate drives would not have benefited from buying them at 2 week intervals.

Meh.

 

[Ericloewe]

My humble opinion:

I prefer Reds over Greens (weren't they supposed to be renamed to Blue? What the hell happened to that story? I wish I'd remember what German publication ran it, so that I can never trust them and their "WD execs" again...) because of TLER (A bad drive is much easier to diagnose if the system didn't crawl to a halt because one drive is trying and trying and trying to read some stupid and trying and trying and trying and trying sector(s) and trying and trying and trying, effectively bringing things to a halt.) and a sane default idle timer.
I fully expect Reds to be Greens with different firmware options selected, with the extra warranty being exclusively a business decision not related to any engineering concerns (except perhaps fewer failures caused by excessive LCCs, if they even consider those worthy of RMA).

As for strategies involving different batches, I think they suffer from two big issues:

• It's hard to get different batches
• Few problems are specific to a small enough batch to make such a plan viable
• Most such issues are easily found during burn-in

 

[solarisguy]

[...] I prefer Reds over Greens (weren't they supposed to be renamed to Blue? [...]

My local store has no Greens in sizes below 5TB anymore, as they are no longer available from their distributor.

 

[danb35]

Greens (weren't they supposed to be renamed to Blue?

WD says so: http://www.wdc.com/en/products/products.aspx?id=770#Tab8

 

[Robert Trevellyan]

WD says so

Ha, I read that as "WD says no", then clicked the link, then I read it again, because, WTF?
:)

 

[Ericloewe]

Oh, lovely, they renamed the Greens and they are still selling the old, 7200RPM Blues.

Sometimes I feel that the whole industry is actively trying to increase consumer confusion (see also: USB 3.1, USB Type C and alternate modes).
Even I am confused, and I spend way more time than is healthy keeping up with this crap!

 

[Arwen]

Back to the original post and subject.

For my 4 x 4TB, I did both spread purchasing AND different models.

1 x WD Red from local store, retail packaging
1 x WD Red from same local store, OEM packaging, bought a few weeks later
1 x WD Red Pro bought from Amazon
1 x WD Red Pro bought from Amazon a few weeks after other Amazon Red Pro purchase

Was even tempted to throw in a different manufacture for 1 or 2 disks.

Some people tell me I am paraniod, (who are they and why are they after me? .

I once had to replace about 30 hard drives in a production data center over a few weeks
due to manufacture defect. In some cases 2 drives in a RAID-5 needed replacement. Had
to wait for 1 to finish, before starting the second. Royal pain in the rear.

So, for my home system, I am designing it to last 5 or more years without serious failure.
(My old Infrant ReadyNAS 1000s lasted at least 6 years... though it was getting slow and
low capacity by today's standards.) Even went so far as to mirror my boot DOM.