ALWAYS close USB flash drive before removing?

On Tue, 17 Nov 2009 13:30:10 -0500, "David W. Hodgins"
wrote:


Also, using the "Optimized For Quick Removal" option increases the
number of small writes to the drive, which will speed up the wearing
out of the drive. Flash drives have a limited number of writes.

Regards, Dave Hodgins

Also, optimized for quick removal doesn't necessarily
increase the number of small writes, a block written at a
time is far smaller than the always present read caching
windows does regardless of the write caching. It's not the
source of the data nor the main memory that is the
bottleneck, it's the USB bus and flash drive's own write
performance, the next write is always waiting whether you
have optimized for quick removal set or not.

What the setting really does is virtualize a completion of
the process so the system can go on to do more work with the
filesystem as if it were finished.

kony wrote:
On Wed, 18 Nov 2009 09:24:48 +0100, David Brown
wrote:


So you should /always/ use safe removal, unless you haven't written
anything to the disk.
Thats just plain wrong. If you have it optimized for quick removal
and its been a while since the write, its fine to unplug it without
safe removal.

That is true, but how long is "a while"?

Several seconds with no I/O indicated.


It is highly dependent on the
size of the write(s), the type of flash disk, and the speed of the
transfer.

None of these are a factor as you aren't timing it then
guessing when it's done. The two indictors when caching is
disabled are that the copy or move dialog window has
finished and the LED has remained indicative of no further
I/O.


That's true enough. The copy/move box is a fairly good indicator, if
all the caching is disabled - the final flushing to disk should be
finished within a couple of seconds of the box closing.

I'm not suggesting it is not generally safe to remove the disk a few
seconds after you expect the copy to be finished - in the great majority
of cases, it /is/ safe. I'm simply saying that there are circumstances
that could make it unsafe. Perhaps /you/ have write caching disabled
and "optimise for quick removal", but not everyone does - it is not
uncommon to use a flash disk to files from someone else's machine.

Using "safe removal" is a good habit, works everywhere, costs nothing,
and avoids accidents on the rare occasions when there might be extra
delays. This is advice to the OP (and others) who would like a nice
general rule. Those who know what they are doing will obviously use
"safe removal" as and when they see fit.

Some flash disks have an LED or an indicator of some sort
that can help, but most don't.

Most do have an LED and the only one I'd ever buy that
didn't, only lacks one because it sacrifices everything to
be very tiny. So yes there is a complication without an
LED, but if someone wants the ability to visually see when
I/O has stopped, the LED is a desirable feature to seek at
time of product selection.


I agree that an LED is nice, but in practice I have seen very few flash
sticks with one. Cost is the main reason few sticks have an LED, rather
than size, but either way they are rare in my experience. You certainly
can't rely on having one on your "average" stick.


Given the O/P's question, I think it's safer to recommend always using
safe removal, even if it is a little more paranoid than "use safe
removal or wait a while".

Safer yes, but since data with any value should never be
kept on only one storage medium, the risk isn't much to be
concerned about.

While you and I know that, there are lots of users who never seem to
understand about backups and extra copies of data. It is certainly not
uncommon for "average" users to copy files to a flash stick as their
only backup. Thus the good rule of thumb for "average" users is the
extra safe one of always using "safe removal".

In comp.sys.ibm.pc.hardware.storage kony wrote:
On Tue, 17 Nov 2009 13:30:10 -0500, "David W. Hodgins"
wrote:

On Tue, 17 Nov 2009 11:15:19 -0500, Eddie wrote:

Or does the emptying of the write cache depend on OTHER FACTORS than
elapsed time?

Suppose the antivirus program is downloading a large update, that
you are not yet aware of. The writes to the flash drive may, or
may not be completed quickly. The only way to be sure, is to use
the remove safely option.

BUT, are you really, REALLY sure? Or are you only placing
confidence in windows doing what it is supposed to do? It's
that a bit of the same thing as assuming the drive LED
access indicator is doing what it is supposed to do?

Windows will tell you that it is safe to remove the drive only
after it has a) written all buffered data and b) blocked the
device so no new writes can happen. In the absence of a programming
error, this is reliable.

Some days I'd trust an LED hard wired to a chip not running
windows, more than windows itself.

The only time it's guaranteed to be safe without that, is if you
haven't written anything to the drive. Your friend is correct
that you should always use the safely remove. Even if 99% of the
time it works ok without it, it's his drive that could effectively
be erased due to fat corruption.


For the sake of argument, let's assume that 99% is a
fictional number and count the ways that a pending write
could be cached still but the user unaware of this.

1) User is unfamiliar with the specific flash drive,
doesn't know what the blinking LED means.

2) Owner didn't pick a drive that has an LED to indicate
I/O or didn't wait a few seconds to be sure there was no
activity.

3) Owner never looked at the LED to confirm it works
properly, there's always that one drive in the whole case
that might be defective.

Well, a defective flash drive may lead to corrutpion anywaus.
And this is a real problem: I have had two defective ones, one
in a long therm overwrite experiment and one as a recovery
boot medium and both have had silent corruption of data, i.e.
no error message at all. This measn their reliability is
significantly below that of a floppy disk.

I currently have a second drive in the torture chair and
will do a writeuo if that gets silent corruption as well.


Also, using the "Optimized For Quick Removal" option increases the
number of small writes to the drive, which will speed up the wearing
out of the drive. Flash drives have a limited number of writes.

Regards, Dave Hodgins

It's not worth considering. You won't wear out a USB flash
drive making backups or moving files around until long after
it's viable lifespan is over. Not even close.

In ordinary use no. I have tortured one Kingston 2GB drive
to death, which took about 3400 complete overwrites or
in the order of a month of continued writing. The problem
was that it then gave wrong data but no error messages. Seems
there is no CRC in these devices or it is switched off.

Arno

--
Arno Wagner, Dr. sc. techn., Dipl. Inform., CISSP -- Email:
GnuPG: ID: 1E25338F FP: 0C30 5782 9D93 F785 E79C 0296 797F 6B50 1E25 338F
----
Cuddly UI's are the manifestation of wishful thinking. -- Dylan Evans

My friends get upset if I don't use the Safe Removal process when
their USB memory is in my PC. Maybe they're repeating a mistake? Or
is there a reason which I'm missing that says I should always use
Safe Removal?

Follow standard procedure.

--
@~@ Might, Courage, Vision, SINCERITY.
/ v \ Simplicity is Beauty! May the Force and Farce be with you!
/( _ )\ (x86_64 Ubuntu 9.10) Linux 2.6.31.6
^ ^ 20:41:01 up 5 days 9:46 1 user load average: 1.00 1.02 1.00
�借貸! ��騙! ��交! �打交! �打劫! �自殺! 請考慮綜� (CSSA):
http://www.swd.gov.hk/tc/index/site_...sub_addressesa

Rod Speed wrote:
David Brown wrote
Rod Speed wrote
David Brown wrote
Eddie wrote

Do I always need to close or stop a USB flash drive before I remove it?

I'm using XP. In XP's Device Manager there is an entry for Disk Drives and when I look at the Policies tab for the
flash drive, it says:

"Optimized For Quick Removal. This setting enables write caching
on the disk and in Windows, so you can disconnect this device
without using the Safe Removal icon."
My friends get upset if I don't use the Safe Removal process when
their USB memory is in my PC. Maybe they're repeating a mistake?

Or is there a reason which I'm missing that says I should always
use Safe Removal?

The main problem with removing USB drives without "safe removal"
(or umount on Linux) is that the OS could still have data in its
write buffers. If you remove the flash drive before the write is
complete, you will lose data, corrupting the file that is being
written, and possibly corrupting the FAT (and therefore lots more
of the flash disk). The "optimize for quick removal" makes windows
save as little in the write buffers as possible, minimising your
risks, but you are still risking data loss and file corruption.
The other choice is to store lots more in the write buffers - that
makes the flash disk far faster for writing, especially for writing
lots of small files, but means greater risk of corruption if you
don't remove it safely.

So you should /always/ use safe removal, unless you haven't written anything to the disk.

Thats just plain wrong. If you have it optimized for quick removal and its been a while since the write, its fine to
unplug it without safe removal.

That is true, but how long is "a while"?

Irrelevant to your stupid /always/ claim.

So how long is "a while"? If you can't quantify it you're just talking crap.

Eddie wrote
kony wrote
Eddie wrote:

Do I always need to close or stop a USB flash drive before I remove it?

I'm using XP. In XP's Device Manager there is an entry for Disk
Drives and when I look at the Policies tab for the flash drive, it says:

"Optimized For Quick Removal. This setting enables
write caching on the disk and in Windows, so you can
disconnect this device without using the Safe Removal
icon."

My friends get upset if I don't use the Safe Removal process
when their USB memory is in my PC. Maybe they're repeating a
mistake? Or is there a reason which I'm missing that says I
should always use Safe Removal?

All safe removal really does is wait to tell you it
can be removed if the buffer is not written out yet.

You say "All safe removal really does is wait to tell you
it can be removed if the buffer is not written out yet."

I thought the Safe Removal window actually *forced*
the data to be flushed out of the write cache?

It does both.

Are you saying it is just a way of making the user wait
until flushing is done (but not start flushing any sooner)?

If that is what he is saying, he is wrong.

As for the optimized for quick removal setting, in use the
main difference is that this keeps the "copying..." visual
indicator window on screen until the copying is done, while
if you chose the other option the copying indicator window
will finish suggesting the copying is finished while all
that is necessarily done is all the data to be copied is now
in the windows buffer.

While writes can be delayed it is not by much, it would take
a fairly heavy use of the system to really delay by more
than the physical write speed and USB bus limitations cause.

IF your drive access LED works properly and you are familiar
with it, you can judge when writing is done by simply
observing several seconds of no activity, but certainly the
safest is to use the safely remove hardware feature.

The best thing to do is just respect your friend's wishes
and use it, regardless of whether it really matters... you
can't spend your life arguing with or educating people about
every little thing... and some of them will get offended.

Many times Safe Removal window says something like operatiom
failed to stop the drive or adaptor. Repeated tries don't seem to
improve this.

On one system I do occasionally get it saying that it cant currently be removed and another go allows it to be removed.

It never keeps saying that it cant be removed.

And I use that system every couple of days.

David Brown wrote:
kony wrote:
On Wed, 18 Nov 2009 09:24:48 +0100, David Brown
wrote:


So you should /always/ use safe removal, unless you haven't
written anything to the disk.
Thats just plain wrong. If you have it optimized for quick removal
and its been a while since the write, its fine to unplug it without
safe removal.

That is true, but how long is "a while"?

Several seconds with no I/O indicated.


It is highly dependent on the
size of the write(s), the type of flash disk, and the speed of the
transfer.

None of these are a factor as you aren't timing it then
guessing when it's done. The two indictors when caching is
disabled are that the copy or move dialog window has
finished and the LED has remained indicative of no further
I/O.


That's true enough. The copy/move box is a fairly good indicator, if
all the caching is disabled - the final flushing to disk should be
finished within a couple of seconds of the box closing.

I'm not suggesting it is not generally safe to remove the disk a few
seconds after you expect the copy to be finished - in the great
majority of cases, it /is/ safe. I'm simply saying that there are
circumstances that could make it unsafe. Perhaps /you/ have write
caching disabled and "optimise for quick removal", but not everyone
does - it is not uncommon to use a flash disk to files from someone
else's machine.
Using "safe removal" is a good habit, works everywhere, costs nothing,
and avoids accidents on the rare occasions when there might be extra
delays. This is advice to the OP (and others) who would like a nice
general rule. Those who know what they are doing will obviously use
"safe removal" as and when they see fit.

Some flash disks have an LED or an indicator of some sort
that can help, but most don't.

Most do have an LED and the only one I'd ever buy that
didn't, only lacks one because it sacrifices everything to
be very tiny. So yes there is a complication without an
LED, but if someone wants the ability to visually see when
I/O has stopped, the LED is a desirable feature to seek at
time of product selection.

I agree that an LED is nice, but in practice I have seen very few flash sticks with one.

Then you need to get out more. Most of us get the exact
opposite effect, very few flash sticks that dont have one.

Cost is the main reason few sticks have an LED, rather than size, but either way they are rare in my experience.

Then you need to get out more. Most of us get the exact
opposite effect, very few flash sticks that dont have one.

You certainly can't rely on having one on your "average" stick.

You dont need to 'rely' on anything, just do it differently if there is no led.

Given the O/P's question, I think it's safer to recommend always
using safe removal, even if it is a little more paranoid than "use
safe removal or wait a while".

Safer yes, but since data with any value should never be
kept on only one storage medium, the risk isn't much to be
concerned about.

While you and I know that, there are lots of users who never seem to
understand about backups and extra copies of data. It is certainly
not uncommon for "average" users to copy files to a flash stick as
their only backup. Thus the good rule of thumb for "average" users
is the extra safe one of always using "safe removal".

Nothing like your original stupidity.

spodosaurus wrote:
Rod Speed wrote:
David Brown wrote
Rod Speed wrote
David Brown wrote
Eddie wrote

Do I always need to close or stop a USB flash drive before I
remove it?

I'm using XP. In XP's Device Manager there is an entry for Disk
Drives and when I look at the Policies tab for the flash drive,
it says:

"Optimized For Quick Removal. This setting enables write caching
on the disk and in Windows, so you can disconnect this device
without using the Safe Removal icon."
My friends get upset if I don't use the Safe Removal process when
their USB memory is in my PC. Maybe they're repeating a mistake?

Or is there a reason which I'm missing that says I should always
use Safe Removal?

The main problem with removing USB drives without "safe removal"
(or umount on Linux) is that the OS could still have data in its
write buffers. If you remove the flash drive before the write is
complete, you will lose data, corrupting the file that is being
written, and possibly corrupting the FAT (and therefore lots more
of the flash disk). The "optimize for quick removal" makes windows
save as little in the write buffers as possible, minimising your
risks, but you are still risking data loss and file corruption.
The other choice is to store lots more in the write buffers - that
makes the flash disk far faster for writing, especially for
writing lots of small files, but means greater risk of corruption if you
don't remove it safely.

So you should /always/ use safe removal, unless you haven't
written anything to the disk.

Thats just plain wrong. If you have it optimized for quick removal
and its been a while since the write, its fine to unplug it
without safe removal.

That is true, but how long is "a while"?

Irrelevant to your stupid /always/ claim.

So how long is "a while"?

Few seconds of no led activity and the Win transfer window gone.

reams of your puerile **** any 2 year old could leave for dead flushed where it belongs

On 18 Nov 2009 10:47:56 GMT, Arno wrote:

In comp.sys.ibm.pc.hardware.storage kony wrote:
On Tue, 17 Nov 2009 13:30:10 -0500, "David W. Hodgins"
wrote:

On Tue, 17 Nov 2009 11:15:19 -0500, Eddie wrote:

Or does the emptying of the write cache depend on OTHER FACTORS than
elapsed time?

Suppose the antivirus program is downloading a large update, that
you are not yet aware of. The writes to the flash drive may, or
may not be completed quickly. The only way to be sure, is to use
the remove safely option.

BUT, are you really, REALLY sure? Or are you only placing
confidence in windows doing what it is supposed to do? It's
that a bit of the same thing as assuming the drive LED
access indicator is doing what it is supposed to do?

Windows will tell you that it is safe to remove the drive only
after it has a) written all buffered data and b) blocked the
device so no new writes can happen. In the absence of a programming
error, this is reliable.

^ That last sentence is the key... of course windows has
never had programming errors and never will, they're design
/features/.



In ordinary use no. I have tortured one Kingston 2GB drive
to death, which took about 3400 complete overwrites or
in the order of a month of continued writing. The problem
was that it then gave wrong data but no error messages. Seems
there is no CRC in these devices or it is switched off.

There might have been a time when one could try and make one
go bad, but today the capacities are much higher so not only
can one select a drive of far higher capacity than they fill
per a given storage requirement like 2GB, it takes a lot
longer to fill it.

That situation may change with the move to USB3 based flash
drives used for other things than backups, if someone buys
into window's readyboost or equivalent future features but I
discount this factor because main system memory has also
become so inexpensive that quite a lot more can be cached in
main memory these days.

On Wed, 18 Nov 2009 09:15:30 GMT, Eddie
wrote:


All safe removal really does is wait to tell you it can be
removed if the buffer is not written out yet.


You say "All safe removal really does is wait to tell you it can
be removed if the buffer is not written out yet."

I thought the Safe Removal window actually *forced* the data to
be flushed out of the write cache?

The write cache was still being written out continuously,
it's not like it waits to write until you or the OS tell it
to.

Are you saying it is just a way of making the user wait until
flushing is done (but not start flushing any sooner)?

Yes, but there are other factors like if you had a file in
use on the flash drive volume, so the safety in removing it
depends on, like most things in life, whether the user is
aware of the details.


Many times Safe Removal window says something like operatiom
failed to stop the drive or adaptor. Repeated tries don't seem to
improve this.

Finding cause would have to be done on a case by case basis.
Sometimes an app won't release a lock on a file, windows
thinks it's still being used. As you noted, using Safe
Remove or not doesn't make a difference then.