Why hard drives fail

[SuperSparks]

A very interesting report reflecting on Google's experience of hard drive reliability is here (PDF):

Why hard disks fail

[Philgo]

Exactly what I was thinking

[smurfy]

As is common in server-class deployments, the disks
were powered on, spinning, and generally in service for
essentially all of their recorded life. They were deployed
in rack-mounted servers and housed in professionally managed
datacenter facilities.

Which therefore eliminates this as reliable analysis for consumer use. Not many home computers are located in a datacentre-standard environment where temperature and humidity and airborne particulates are strictly controlled.
When the test environment doesn't really go below 15 or above 50, it's not replicating real-world consumer use.
So I doubt their conclusion of "temperature not as much of a factor as first thought" is strictly correct for aanything other than controlled environments.
Also the "always on" nature eliminates one of the factors that usually invokes a debate - does regular powering on and off of computers increase the stress on the components and increase the likelihood of failure?

Still interesting reading though.

[JPnyc]

I've yet to have a disc fail on me. I do tend to leave the desktop on more often than I shut it down, but with laptops that's a bit difficult to do.

[Philgo]

I've yet to have a disc fail on me.

Famous last words

[Train]

Famous last words

You BET!

[orc]

I've have 33 dead disks in a box, that represents how many I've changed in the last year. I have two 90 litre garbage cans full of parts that I take to hazardous recycling once a year. I didn't look at the pdf, but if it's the one I've seen, it says failure rates are 2-4%. I sell systems, and I'd say that's right on. If you don't have two backups of your data, it's data you don't need.

[JPnyc]

33 discs in a yr? What the...? I've run PCs daily for 8 yrs without changing a single piece of hardware.

[Tuttle]

I suspect those are from customers' systems.

But I want JPnyc's record. I've had one drive die on me here at home (previous PC), and my PC at work has lost four drives in its two years of life. We had one server rack with 28 drives that was losing one every week or two, but that turned out to be zinc fibres growing out of the floor tiles and getting blown into the hardware. That problem went away once we replaced the floor.

[JPnyc]

Were these "always on" systems? I've heard lots of people say that powering on and off shortens the life of components, but I've yet to see hard evidence of it. I rarely kept the PC on overnight, and as I said that one last 8 yrs. My laptop is at least 6 yrs old.

[crunchie]

Hard drives don't fail, do they

[Allthumbs]

My Dell Dimension XPS H266 is ten years old this year. I power up about 3 times a day (usually) and still on the same (Maxtor) hard drive. I would think a computer that is on 24/7 would have bearing trouble. Do you leave your car run all the time? The last I heard, a hard drive MTFB is about 5,500 hours of operation.

[ua549]

My notebook IDE drives consistently failed between 2 weeks and 60 days. It did not matter what the brand of hard drive was. The notebooks were various brands - Toshiba, IBM and HP. The notebooks ran 24/7/365 at the high end of the permitted environment of 90°F. IMO the drives failed from the heat.

I now have Core2 Duo notebooks that run about 40°C cooler than the P4 Mobile notebooks. So far the SATA drive, a Hitachi, has run for 4 months and has only encountered a few bad sectors. Chkdsk is run twice a day.

All of my non-notebook drives are SCSI. Some are over 20 years old and I have not lost one yet.

[JPnyc]

Yeah the core 2 duo is what I have now also.

[Tuttle]

My home PC is 8-12 hours a day. All the stuff at work is 24/7.