Heat doesn't kill hard drives. Here's what does

[U2gamer]

like that win already lor... our HDD will always fail faster than other people. Begins another loop of back up and re-back up...

Sian. Have to back up data in multiple hdds. Then like that the number of years of warranty also very important. If fail then can rma. If not need to be another one.

Posted with HardwareZone Forums app

 

[Rogue]

Relative humidity is based on temperature.... Higher temp-> More water vapor stored in a certain volume of air under a certain amount of pressure.

Even the other thread, there's some misconception.

 

[jtjt00]

This thread should be merged into here:
http://forums.hardwarezone.com.sg/hardware-clinic-2/heat-doesnt-kill-hard-drives-heres-what-does-5318758.html

 

[xiaowhois]

This thread should be merged into here:
http://forums.hardwarezone.com.sg/h...kill-hard-drives-heres-what-does-5318758.html

high chance is that thread kanna merged here

 

[koreanlover]

So HDDs fail slower in aircon rooms and temperate countries?

 

[jtjt00]

Relative humidity is based on temperature.... Higher temp-> More water vapor stored in a certain volume of air under a certain amount of pressure.

Even the other thread, there's some misconception.

It is not true that cold air is always dry as in the case of temperate countries. In my work to calibrate Hisamatsu TN-2500 thermo-hygrograh, found that test floor AHU cold air is high in relative humidity 70-80% be it from ceiling or under raised floor. Only when machines are running, will the temperature rise and %RH drop.

As such, servers on rack at the bottom and with HDD placed in front exposed to high humidity. On the other side of the scale, humidity < 20%RH present the ESD damage threat. This is mentioned in the article.

 

[kandinsky]

So HDDs fail slower in aircon rooms and temperate countries?

They found that the data centres classified as 'Hot and Humid' had higher failure rates.

Data sources. We collect data from nine hyperscale Microsoft
datacenters spread around the world for periods
from 1.5 to 4 years. The data includes component health
and failure reports, traces of environmental conditions,
traces of component utilizations, cooling energy data,
and asset information.

The datacenters use a variety of cooling infrastructures,
exhibiting different environmental conditions,
hardware component reliabilities, energy efficiencies,
and costs. The three first columns from the left of Table 2
show each datacenter’s tag, its cooling technology, and
the length of data we have for it. The tags correspond to
the environmental conditions inside the datacenters (see
caption for details), not their cooling technology or location.
We classify a datacenter as “hot” (“H” as the
first letter of its tag) if at least 10% of its internal temperatures
over a year are above 24◦C, whereas we classify
it as “humid” (“H” as the second letter of the tag)
if at least 10% of its internal relative humidities over a
year are above 60%.

Prior works had shown that disk failures are
the most common for stable environmental conditions,
e.g. [27]. Our data shows that they also dominate in
modern, hotter and more humid datacenters. Interestingly,
as we discuss in Section 7, the placement of the
components (e.g., disks) inside each server affects their
failure rates, since the temperature and relative humidity
vary as air flows through the server.

Disk failure rates. Table 3 presents the disk AFRs for
the datacenters we study, and how much they differ relative
to the AFR of one of the datacenters with stable temperature
and relative humidity at each spatial spot (CD1).

We make several observations from these data:
1. The datacenters with consistently or frequently dry
internal environments exhibit the lowest AFRs, regardless
of cooling technologies. For example, CD1
and HD1 keep relative humidities stable at 50%.
2. High internal relative humidity increases AFRs by
107% (HH1) to 260% (HH4), compared to CD1.
Compared to HD1, the increases range from 55% to
170%. For example, HH1 exhibits a wide range of
relative humidities, with a large percentage of them
higher than 50% (Figure 1).
3. Free cooling does not necessarily lead to high AFR,
as CD3 shows. Depending on the local climate (and
with careful humidity control), free-cooled datacenters
can have AFRs as low as those of chiller-based
and water-side economized datacenters.
4. High internal temperature does not directly correlate
to the high range of AFRs (greater than 3%), as
suggested by datacenters HD1 and HD2.
The first two of these observations are indications that
relative humidity may have a significant impact on disk
failures. We cannot state a stronger result based solely on
Table 3, because there are many differences between the
datacenters, their servers and environmental conditions.
Thus, in the next few pages, we provide more detailed
evidence that consistently points in the same direction.

http://0b4af6cdc2f0c5998459-c0245c5...ackcdn.com/23104-fast16-papers-manousakis.pdf

Quite interesting read, a lot of stuff catch no ball (all the formulas ), but some stuff is in plain english.

 

[jtjt00]

Temperature & relative humidity is never uniform across a given aircon room, there are areas with low/high temperature/%RH. And the wall thermo-hygrograph only record a single point to satisfy the auditors.

 

[commach]

high chance is that thread kanna merged here

No, merged according to date of creation, HWZ administrator's thread is not "untouchable", proper house keeper is the way to go.