Google article on Disk Failure's

[Mike32]

Thought you guys might be intereseted in this:

http://labs.google.com/papers/disk_failures.pdf

Mike

 

[abcdelight]

I just read that at /. and it was very interesting. I liked the part about heat and start-up cycles not having a high failure correlation. There has always been that camp that says "don't" turn off that hard-drive too often because it streses it. I never felt in sync with that idea. I notice they didn't which HD maker did best (they did imply there was one but then didn't name them).

 

[yaric]

Cool, the part about the temperature not mattering much makes sense. But thermal cycling will put any machine through stress. That's why semi engines last for millions of miles, they are constantly running, few thermal cycles. Normal cars are started up and shut off all the time and therefore have a much higher failure rate. They didn't really talk about power up/thermal cycles in this as they said all theres were on almost continuosly...

So i guess you can say i'm part of the camp that says "don't turn your hds off".

But actually i'm not against turning them off. I'm just against turning them off and letting them cool, then turning them back on all the time. Thermal cycling is bad! especially when tolerances are as close are they are in a hd, the metals heat up and expand, then cool down and contract.

 

[Met-AL]

That's why semi engines last for millions of miles, they are constantly running, few thermal cycles. Normal cars are started up and shut off all the time and therefore have a much higher failure rate.

No they last for millions of miles because they are huge, over engineered, run on diesel fuel and run cooler, and most importantly, they spin slower ~2000 RPM. In fact, you will find a huge push in the trucking industry away from your exact line of thinking. Trucks no longer spend hours idling, but instead the engines are shut off at every opportunity to conserve fuel which adds up to millions of barrels of oil each year and the money to go with it.


But anyways, the findings by Google are interesting. It sure shoots a big hole in the bored "enthusiast's" theory that we all need to cool this and cool that and leave this running forever cause turning it off shortens it's life. But, there will be people that will blindly argue the findings to death, and there will be others that say..hmm, my hard disk isn't gonna fail because I didn't give a shit enough to rubberband a fan in my case to blow air on my hard disks. Maybe those MTBF numbers do add up and leaving your hard disk on for 6yrs straight is no better then if you would have turned off your PC for the 90% of the time you weren't using it and maybe saved a few hundred bucks in electricity.

 

[drizzt81]

No they last for millions of miles because they are huge, over engineered, run on diesel fuel and run cooler, and most importantly, they spin slower ~2000 RPM. In fact, you will find a huge push in the trucking industry away from your exact line of thinking. Trucks no longer spend hours idling, but instead the engines are shut off at every opportunity to conserve fuel which adds up to millions of barrels of oil each year and the money to go with it.

It is likely that there is a minimum in the cost function C(s,i), where s is the number of restarts and i is the time spent idling. Given the cost and energy expenditure to produce a semi engine and the cost of gasoline and rate of consumptions during idle time, there will likely be a time t, for which idling either longer or shorter will increase the cost.

 

[Met-AL]

It is likely that there is a minimum in the cost function C(s,i), where s is the number of restarts and i is the time spent idling. Given the cost and energy expenditure to produce a semi engine and the cost of gasoline and rate of consumptions during idle time, there will likely be a time t, for which idling either longer or shorter will increase the cost.

Exactly.