Is "USB-drive time bomb" for real?

[Danny]

Is this story (from 2005) on the Dan's Data web site correct? It says
ext drives will burn out faster because they don't spin down.

http://www.dansdata.com/gz055.htm

All things being equal, I would have guessed start/stop wears out
drives more than the continuous running Dan is worried about.

Any info?

 

[Bob Willard]

Is this story (from 2005) on the Dan's Data web site correct? It says
ext drives will burn out faster because they don't spin down.

http://www.dansdata.com/gz055.htm

All things being equal, I would have guessed start/stop wears out
drives more than the continuous running Dan is worried about.

Any info?

Depends on your workload, and on the HD. For instance, I run an app
that writes a log entry to the HD every 30 minutes. So, for a HD that
is rated for 50,000 start/stop cycles and has a MTBF of 1,000,000 hours
(rather typical for a decent HD), letting it spin down after 15 mins of
inactivity will bring about EOL in <3 years, while avoiding spindown
will defer EOL for ~114 years.

No, I don't believe those absolute numbers either. But the relative
numbers clearly dictate my management practice.

As a practical matter, many HDs in SOHO usage die due to either:
1. poor airflow => poor cooling => over-temp => early death
2. finite size incapable of growing to match software bloat

 

[Arno]

Is this story (from 2005) on the Dan's Data web site correct? It says
ext drives will burn out faster because they don't spin down.

All things being equal, I would have guessed start/stop wears out
drives more than the continuous running Dan is worried about.

Any info?

Quite frankly the article sounds like rambling to me.

With Fluid Dynamic Bearings (standard today) the drive
does not wear out while spinning. For traditional
ball-bearings this is different.

There are limits on how many spin-up cycles you can
have. The numbers for notebook drives are about 500'000,
while desktop drives have about 1/10 of that.

The only thing left for continuously spinning drives is
heat. Heat kills over time. However, drives have become
much lower in power consumption, so this is not
an issue for a well-cooled HDD either.

Arno

 

[Arno]

Depends on your workload, and on the HD. For instance, I run an app
that writes a log entry to the HD every 30 minutes. So, for a HD that
is rated for 50,000 start/stop cycles and has a MTBF of 1,000,000 hours
(rather typical for a decent HD), letting it spin down after 15 mins of
inactivity will bring about EOL in <3 years, while avoiding spindown
will defer EOL for ~114 years.

No, I don't believe those absolute numbers either. But the relative
numbers clearly dictate my management practice.

As a practical matter, many HDs in SOHO usage die due to either:
1. poor airflow => poor cooling => over-temp => early death

Agreed. And this may kill the electronics even more often
than the mechanics.

2. finite size incapable of growing to match software bloat

Agreed as well. Not dead, but useless. Not so far off.

Arno

 

[Rod Speed]

Is this story (from 2005) on the Dan's Data web site correct?
Nope.

It says ext drives will burn out faster because they don't spin down.

Thats just plain wrong. I choose not to spin down
my internal drives and have never ever had one fail.

And I leave all my systems except the battery powered ones on all the time too.

http://www.dansdata.com/gz055.htm

All things being equal, I would have guessed start/stop wears
out drives more than the continuous running Dan is worried about.

In practice as long as you dont get too carried away and spin them
down every minute etc, neither approach wears out the drive.

Even XP and Vista dont spin down internal drives anymore.

 

[Franc Zabkar]

Depends on your workload, and on the HD. For instance, I run an app
that writes a log entry to the HD every 30 minutes. So, for a HD that
is rated for 50,000 start/stop cycles and has a MTBF of 1,000,000 hours
(rather typical for a decent HD), letting it spin down after 15 mins of
inactivity will bring about EOL in <3 years, while avoiding spindown
will defer EOL for ~114 years.

No, I don't believe those absolute numbers either. But the relative
numbers clearly dictate my management practice.

MTBF describes failure rates for drive populations, not individual
drives:
http://www.hitachigst.com/hddt/know...a03ca96b145bc68e86256df6004d318e?OpenDocument

===================================================================
MTBF is a statistically derived number, typically based on ideal
environmental conditions, under which one would expect a failure.

For example:

A product has a MTBF specification of 1.2 million hours.

This means that in a specified environment with a population of 1.2
million (drives) one could expect a failure every hour (from within
that population)
===================================================================

As a practical matter, many HDs in SOHO usage die due to either:
1. poor airflow => poor cooling => over-temp => early death
2. finite size incapable of growing to match software bloat

- Franc Zabkar

 

[Arno]

MTBF describes failure rates for drive populations, not individual
drives:
http://www.hitachigst.com/hddt/know...a03ca96b145bc68e86256df6004d318e?OpenDocument

===================================================================
MTBF is a statistically derived number, typically based on ideal
environmental conditions, under which one would expect a failure.

For example:

A product has a MTBF specification of 1.2 million hours.

This means that in a specified environment with a population of 1.2
million (drives) one could expect a failure every hour (from within
that population)
===================================================================

There is a second way to look at it: During the component life
(usually 5 years), the MTBF gives you a failure probability per
time number for a drive.

A MTBF of 1.2 million hours just means a failure probability
of 1/(12 million) per hour. After the component life is over,
the MTBF is no longer accurate.

So, yes, there is no relation between the lifetime of a drive
and its MTBF, as the MTBF only holds during the component life.

Arno