I read it on the internet, so it must be true!
Hi Zap,
there is a graph in the Google Hard Drive reliability paper that shows an inflection point for failures starting below 50 degrees C; that is what I was referencing in opining that 50 degrees C was too hot.
Here is that graph:
http://storagemojo.com/wp-cont.../afr_temp_age_dist.png
Not many Google drives run over 40-45C or so....50C is definitely way up there, and, look at that failure rate curve. (Note: about 40C, or just about human temp, seems to be good....) So, cooling his drives down 10C should, in theory, help their longevity, according to the Google study. (By as much as 50% per EE equation below....)
My drives, two Seagates in a Supermicro case with great cooling, quiet temp controled 120 mm fans: 82 degrees F/ about 28 C.....And yes, I see that at the temp where my drives run, the failure rate is about the same as at 50C....
From WD website for WD640 SE16:
http://www.wdc.com/en/products/Products.asp?DriveID=394
Cool - Keeping the drive cool enhances reliability. WD continues to develop new and innovative ways to keep drives cool while they are operating.
Operating temp:
Temperature (Metric)
Operating 5° C to 60° C
Those are the
limits, not their recommended operating environment. (In fact, the Google article said hard drives that are too cold fail faster than the too hot drives...and no, I did not miss that....) The operating limits for humans goes from pretty low to pretty high as well, but, I am not very efficient or reliable at 50C, and I have been there....
From:
http://arstechnica.com/news.ars/post/20070225-8917.html
When it comes to the question of temperature and drive failure, though, Gibson says that the "very high temperatures" which can affect drive longevity are actually <quite common in home PCs. "SpinRite will stop mid-run when drives become too hot," he says, "but SpinRite will only stop when the drive gets really hot." And Gibson knows from user reports that his tool does regularly stop to let drives cool down. "So this informs us that these 'extreme' temperatures are actually being encountered in the real world and are limiting drive lifetimes."
Anyway, from looking at the Google article graph, *and* from reviewing EE equation below, I do recommend TC cools them down. (Note: I will not turn you into a straw man and claim you recommended he heat them up to 60C.....140F, hotter than McDonald's coffee....)
You did get me thinking a bit about this: hard drives are both electronic and mechanical devices (at least for now): I wonder what part of the drive generally fails for Google: the armatures, disk motor, the bearings, some other mechanical part, or the electronics?)
NXIL
Note: and, I realize that if in theory according to this equation you cooled a hard drive with liquid helium to 1 degree K it should last longer than the universe will.....
EE equation:
http://electronicdesign.com/Ar...cleID/16767/16767.html
The relationship between operating temperature and reliability is defined in a system's failure rate (useful system life in failures per 106 hours), as expressed in the Arrhenius Model:
Lambda = A *
e to the power of (Ea / k * T)
where:
? = failure rate
A = constant
Ea = activation energy for the particular failure mechanism
k = Boltzmann's constant
T = Kelvin temperature
Equation 1 shows that failure rate is a function of the temperature stress: the higher the stress, the higher the failure rate (more failures per 106 hours). Typically, according to Equation 1,
each 10°C rise in temperature increases the failure rate by 50%. Conversely, cutting the operating temperature by 10°C reduces the failure rate.