VanguardLH <
V@nguard.lh> wrote:
jerryab wrote:
Filip454 wrote:
This is the biggest myth, which is also constantly repeated on the
internet.
Lower temperature =/= running longer
It is no myth for electronic components. Running hotter = shorter
component life.
Except your hypothesis is one-ended. Running too cold is more deadly
than running too hot. Also, there is a difference between a drive that
is running hot (outside of its operating temperature range) and just
running hotter (which is relative to some base value that is often never specified).
Google and others have actually done massive studies on hard drive
failure rates under various environmental conditions - versus you and
others promulgating that age-old and unfounded myths based on casual observation.
There's nothing unfounded or mythical about it. The "Hardware
Lifetime Models" section of the paper you reference (page 5 in
the PDF) clearly describes the known correlation of increased
temperature with the decreased lifetime of electronic components and
devices.
It's point was that with higher humidities (which are more common in
cooler datacentre enviroments, but are also not an essential
characteristic of them), corrosion beats the other effects of higher temperatures "to the punch" in terms of what kills off HDDs. If you
can achieve cooler operation without higher humidity, then that's
still preferable to a higher temperature.
Now what I don't understand is why there's no consideration given to
preventing the damage that can be caused by humidity in the first
place. Humidity and electronics are long adversaries, especially in
millitary applications. During WWII lacquers to seal electronics
against such harsh enviroments were successfully developed and are
today available cheaply for spray-on application (I've got a can
myself. Oh hell, it's leaked...). It's still applied to most
millitary and other high-reliability electronics, but generally
consumer and business electronics are used in enviroments where
corrosion isn't considered enough of an issue to warrant this
protection. To me this suggests that datacentre design has allowed
the humidity levels to slip a bit above normal, but in any case I
don't see why circuit board laquer (or "conformal coating", to use
the modern industry term) couldn't simply be applied to HDD circuit
boards to prevent the problem with only a minor associated cost.
https://en.wikipedia.org/wiki/Conformal_coating
This wouldn't solve issues with corrosion to the HDD connectors,
which the study included as part of the problem. However again this
is no new issue for the electronics industry, with various approaches
available to combat the problem at a hardware level.
https://www.zdnet.com/article/heat-doesnt-kill-hard-drives-heres-what-does/ (use the link to the PDF to read the research)
https://www.teamsilverback.com/humidity-in-the-data-center-the-real-hard-drive-killer/
Turns out humidity has a greater effect on hard drive longevity than
does temperature (wherein the restriction that you are running the hard
disk within its rated operating temperature range, not at extremes).
Hard disks are not sealed. The cinter filter only prevents particulate debris from getting inside, not gases or vapors.
That's nothing to do with the humidity problem. The original paper
states that humidity causes failures due to corrosion of the
electronics on the HDD controller circuit board, and the HDD
connectors. (see "causes of failure", page 7 in the PDF).
Surge current (electrical stress), thermal stress (temperature delta),
and mechanical stress are more important than running hot or cold (but
again within the rated operating temperature range).
Actually the paper states that temperature is the next most important enviromental factor after humidity (though the effects of the latter
are claimed to be much more significant). (pg. 11, though it's stated
more plainly somewhere else that I can't find now).
USB-attached
drives and even internal ones (depending on power options) can be
stressed more by power cycling them or repeatedly spinning down and
resuming on activity.
Yes, though that's a separate matter to operating
temperature/environment.
Even physical orientation of the disk has more of an effective on
longevity than does runnning at the high end of the operating
temperature range. Manufacturers say there is no difference because
they are specifying within the MTBF rating for their product. Heat
rising off a horizontal platter does not cross over and reheat the
platter as would occur in a vertical orientation. This is the same consideration when planning air flow within the chassis: do not push pre-heated air over components that you are trying to cool. Also, in
the past, manufacturers used to recommend the drive was formatted in the
same orientation under which it would be used. In other words, don't
format when horizontal and then use when vertical. Mostly that has been
lost regarding longevity of the drive but there is still a performance degrade. See:
http://lowendmac.com/2018/does-hard-drive-orientation-affect-performance/
The results show "it depends on the make and model of the drive". So,
it's not a hard and fast rule but can occur.
USB-attached and even internal disks using Power Options will spin down
when idle. This means cooling down. It means a power surge on spin up
which means heating up. Higher activity means more energy consumed
which means more heat; however, activity fluctuates for end user hosts
versus file servers where the drive is likely to be constantly active. Electrical stress (surge), thermal stress, and mechanical stress also
factor into longevity. As the Google study shows, age far outweighs any temperature factor regarding longevity - with 2 years being the major
factor to change from 2% to 8% AFR (annualized failure rate). That is, longevity is not linear with age of the drive.
I haven't read this Google study, but assuming it follows the same
definition of "old age" as the humidity study, this includes things
like "Power-On Hours" and "Spin-up Count", so a "failure" here is
presumably where a drive is simply towards the end of its estimated
lifetime and therefore retired. In this case it's irrelevent for
comparing with other "failures".
--
__ __
#_ < |\| |< _#
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)
