How long do drives last.

[Chewy509]

Saw this linked via Slashdot...

Nice intro article on drive failures...

http://www.extremetech.com/computing/170748-how-long-do-hard-drives-actually-live-for

 

[jtr1962]

It's good to see some data on this, especially the part about wear-out failures. Now I wonder how many hours of running starting and stopping a drive is equivalent to? I tend to let my main desktop run 24/7, mostly on the theory that repeatedly turning it on and off is bad for the electronics, and especially bad for the hard drives. I would love to see data one way or another. If starting and stopping has a negligible effect on wear, then most home PCs which are rarely on more than 12 hours a day will be obsolete long before the hard drives are around long enough to wear out.

I also feel installation has a huge effect on hard drive longevity. I've always been super careful installing hard drives, to the point where I would avoid even a gentle tap. My theory is any impacts will cause some microscopic particles to come lose inside the hard drive. These in turn would accelerate wear. I would imagine in a data center they don't have the time to baby their drives, so the data represents what happens when you install the drives fairly roughly.

I have a 2006 vintage 200 GB drive in my main desktop now (along with a c. 2010 2TB drive and a 1-year old 240 GB Intel SSD). All three have mostly been running 24/7 since I bought them. Extrapolating the data in the article, less than 40% of drives will make it to their seventh year. I wonder if my gentle installation and 24/7 operation have increased the potential drive lifetime substantially?

SSDs should have even longer lifetime. There is no mechanical wearout mechanism. The longevity of flash memory is well known and usually compensated for by wear-leveling mechanisms, to the point that hundreds of TBs of writes are needed to wear out the drive. I wonder if in the final analysis an SSD which isn't worn out by writes will have a longevity similar to other well-designed electronics (i.e. several decades at least)?

 

[mubs]

My systems have been shut down at the end of the day (since 1991). Since about 1999, though, my drives have had a fan in front of them and have always been cool to the touch. I've never had a hard drive fail - even external ones.

 

[Mercutio]

There's a file server in my back bedroom with 36 high capacity 7200rpm drives attached to it. It's shoved in a closet with the door closed and there's not really any moving air. The machine itself is relatively cool (4U case, giant Scythe HSFs on both CPUs), but there's no active cooling at all on any of the drives in either chassis where they're attached. I might lose one drive every nine months or so. I usually do a few full sector-by-sector tests on them before I decide they're acceptable to add to an array and I do occasionally find a dud drive in that time, but I let drives get hot and it doesn't seem to be a problem.

 

[Howell]

I give my drives a good whack before I install them just to make sure any loose sub-atomic particles are off the platters and not going to interfere with the head. I must be getting all of the infant mortality drives with the failure rate I have.

 

[Howell]

An SSD that has a high percentage of writes to a high percentage of the drive is the quickest road to SSD failure. Both user and manufacturer would want the drive to at least reach the end of the warranty. One way to help this is to buy a larger drive than you need and let the wear leveling spread the writes over a larger number of cells. But this is dependant on only filling the drive up to a certain percentage and relying on the consumer. How long before manufacturers are adding more chips just to meet warranty?

 

[Mercutio]

I only beat the ones that say Western Digital on them.
But to be fair, they deserve it.

 

[ddrueding]

How long before manufacturers are adding more chips just to meet warranty?

They already do. Depending on the manufacturer and segment, up to 25%+ extra.

 

[Clocker]

I just had a WD20EARS fail a few months after the warranty expired. Didn't have a lot of use but it was in my file server.... always kept relatively cool. My first real failure.

 

[Chewy509]

Here is a research article by Google Engineers from back on '07, mainly for comparison.

http://research.google.com/archive/disk_failures.pdf

Unfortunately, publicly available peer-reviewed research on large numbers of failures and their conditions is very rare... (even hitting up the IEEE CSDL and ACM libraries doesn't show a big list either).

 

[Chewy509]

Also, based on personal experience I see most failures within 6months, and mostly not until 3+ years... which fits in with their report... (All drives have typically had active cooling in place, and treated well by the technicians).

But in interesting observation made by a friend who works in a small PC store... Most of the drives that experience early death (less than 6 months) all tend to come through a particular supply chain, but the same model drives from a different distributor all tend to last longer... I wonder how many early failures are due to mishandling during initial shipment from the supplier through to the end user?

 

[Howell]

Seems like something that could be discovered with a couple dummy drives per case pack with telemetry.

 

[Tannin]

But in interesting observation made by a friend who works in a small PC store... Most of the drives that experience early death (less than 6 months) all tend to come through a particular supply chain, but the same model drives from a different distributor all tend to last longer... I wonder how many early failures are due to mishandling during initial shipment from the supplier through to the end user?

Agree entirely, Chewy. I first started making that point more than a decade ago, and it's probably almost as true now as it was then.

 

[ddrueding]

Agree entirely, Chewy. I first started making that point more than a decade ago, and it's probably almost as true now as it was then.

I also agree. Which is why I buy in the largest quantities feasible (usually a dozen) from the largest supplier (Amazon) with the fastest shipping. Fewest hands touching it for the smallest amount of time.

 

[Handruin]

But in interesting observation made by a friend who works in a small PC store... Most of the drives that experience early death (less than 6 months) all tend to come through a particular supply chain, but the same model drives from a different distributor all tend to last longer... I wonder how many early failures are due to mishandling during initial shipment from the supplier through to the end user?

Agree entirely, Chewy. I first started making that point more than a decade ago, and it's probably almost as true now as it was then.

I also agree. Which is why I buy in the largest quantities feasible (usually a dozen) from the largest supplier (Amazon) with the fastest shipping. Fewest hands touching it for the smallest amount of time.

I wonder how much of it is the (mis)handling in the supply chain and also the manufacturer's and/or reseller's ability to protect (package) the drive during shipping.

 

[Mercutio]

All I'm going to say is that I seem to have a statistically significant number of additional problems with WD drives vs. anything else I have ever purchased, no matter where bought them or how they were delivered to me. That is the least biased comment I can make about it.

My common practice is to buy at least a dozen drives at a time to ensure I get a proper shipping container. I do think that helps me a lot.

 

[mubs]

Even with WD Retail drives (I mean the single drives in full retail packaging) ?

 

[Chewy509]

Even with WD Retail drives (I mean the single drives in full retail packaging) ?

Most likely even more so, as the shipments of retail boxes are not only to PC shops, but also to the likes of Walmart/Costco/KMart/Target/etc... And we know how well general retailers treat electronic gear during shipment...

 

[mubs]

But the idea is that single retail units are packed so well that you can toss them around and the drive shouldn't get hurt. The "oem" drives are sold wrapped in an anti-static bag and noting else. If you drop that, then all the bits get mixed up inside.

 

[ddrueding]

Perhaps it would still have thermal fluctuation in their crappy warehouses to deal with? Who knows. I just want as few people as possible to touch the drive as little as possible from the factory to me.

 

[timwhit]

Maybe someone should plan a trip to Thailand and pick us all up a bunch of drivers?

 

[Mercutio]

But the idea is that single retail units are packed so well that you can toss them around and the drive shouldn't get hurt. The "oem" drives are sold wrapped in an anti-static bag and noting else. If you drop that, then all the bits get mixed up inside.

"Single" drives are packed in cartons of between eight and 24 drives depending on the manufacturer, at least as far as shipping to the local vendor. The interior of the carton is custom foam of some sort so there's very little room for the drives to move around. The singles are pulled from one of those cartons for individual sale, but at least until they get that far they're in what is probably as close to ideal packaging as anyone could hope to get.

 

[mubs]

I've been talking about one of these.