USB thumb drives.

Jimmy Wilkinson Knife wrote:
On Tue, 22 May 2018 08:20:46 +0100, Lucifer Morningstar

SSDs are not used in servers due to their unreliability.

They are MORE reliable than rotating rust disks.

I they're not used in servers it's because they can't stand the higher
amount of writing. And that depends on what you're doing with the
server. If i made a server where I wanted fast disk access, but it
wasn't written in huge quantities, I'd use SSDs.

https://ark.intel.com/products/97164...#tab-blade-1-0

# Capacity 750 GB
# Endurance Rating (Lifetime Writes) 41000 TBW
# Mean Time Between Failures (MTBF) 2 million hours
# Enhanced Power Loss Data Protection Yes === "I'm an enterprise product"

That drive can be written end to end, about 55000 times,
if you believe the endurance rating.

There are also flash based storage subsystems with internal
RAID, such that if a chip fails, another (unused) chip can
be put in place of it, and rebuilt using redundancy info
in the chip array. I think this is called RAIN.

https://www.micron.com/~/media/docum...f_ssd_rain.pdf

It's possible there are also flash drives (think of a 4U form
factor), where flash chips are arranged on cards, and can
be replaced hot when they fail. Take the top off, plug in a
new card (wait for rebuild). These are products you won't find
on the web, and if you have to ask the price, you can't afford
one. (It's similar to PCI Express, where there are all sorts of
whizzy PCI Express technologies out there, which aren't advertised
on the web. Those might be used in HPC environments.)

Most of the advertising we see, is for the "rubbish stuff" :-)

Paul

On Tue, 22 May 2018 21:16:44 +0100, Paul wrote:

Jimmy Wilkinson Knife wrote:

At what age does the data become unreadable if the drive has not been
powered up?

Flash is generally quoted as holding charge on a floating
gate for around 10 years.

That means, if you're using SSDs for archival storage, they
should be plugged in and re-written every five years, at a guess.

For a given design, I don't know how to guess at that value,
and the 10 year number is merely a "starting point, ball park number".

If I had an SSD today, that was as old as the oldest hard drive
in the room, chances are it would throw a CRC error or two, signifying
the error correction couldn't fix the number of errors accumulated
in a sector.

In my experience, hard drives start making nasty noises and or getting bad sectors well before SSDs show any signs of problems. Back in the days of not much RAM, pagefiles wore the poor things out really quickly.

It's a mirrored array, and the drives have different SMART data, even
though they're identical and were installed together, so one should fail
well before the other and prevent a problem.

Well, I don't want to propose something stupid to you,
and cause the mirror to break as a side effect. You have
to be careful that any soft-raid methods don't "track"
what you do to them, and then the next time you boot
into the "working" configuration, the status is
degraded and it costs you another rebuild. If you move
one of those drives somewhere so you can read the SMART,
you might upset the array status.

I'd mess about if it was a machine with not much on it. I've got 4 PCs doing nothing but science research. An OS and one program to reinstall, no data. But the SSDs in question are my main PC with everything on it. When I built it 5 years ago to replace an older one, it took me weeks to get everything installed and setup the way I wanted it.

If you don't have SMART visibility, and you insist
on running a RAID 1 mirror, I would recommend
to you that you mix drives from different manufacturers.
Pair an Intel branded 512GB drive with a Samsung branded
512GB drive. That should de-correlate things enough, that
there won't be any unfortunate accidents. I personally
would not pair two identical Intel drives in a RAID 1
mirror, if you paid me :-) I'm be a "lucky enough guy"
to have Windows Defender and Search Indexer keep writing
to C: just after the first drive fails, until the second
drive fails and I'm toasted. That's what would happen
to me if I tried that.

I tend to use identical drives in mirrors as I thought otherwise you just degrade it to the speed of the slowest one. I also only use Crucial SSDs for reliability.

Anyway, the SMART I get from Speedfan indicates one of my SSDs will fail a long time before the other. I doubt they're identical enough to fail within a short time period. If I was that worried, maybe I'd use RAID 6.

With mirrored drives *you still need backups*.

Yip, I have what used to be USB3 external drives I swapped around for full backups. But I've now put one into each of the science computers so I can backup over 10Gbit network (into another building, so there's fire protection too). The drives are fast enough, USB3 is fast enough, but the ****ty interface in the external caddies was causing ridiculous bottlenecks and backups took days.

If the 5V rail on your PSU overvolts, and burns both
SSDs at the same time, "you got nuttin".

I use quality PSUs, I've never known even a ****ty one to overvolt (undervolt, yes). Even the cheap ones tend to commit suicide without killing anything else. I've had a couple of cheap ones go bang in the past when I "overloaded" them, without frying anything else. "Overloaded" means daring to draw 500W from an 800W supply. If it's not a big brand name, don't expect more than 50% of the rated watts before an explosion. The decent Corsair ones are wonderful, they don't even power up their own fan until you draw a lot of current - virtually no heat from them at all.

We do backups
to protect against lightning and PSU failures and ransomware.
The mirror idea, isn't "the Space Shuttle". It's not sufficient
redundancy for disaster planning. It's *not* a substitute
for backups.

I know, I never said it was. But it's damn convenient. No reinstalling, no downtime, just shove in another drive with the PC still turned on. I've never had to restore much from backups when using a mirror or any other redundant array. Just files I accidentally deleted. And decent RAID controllers let you specify one or more "spare drives", so as soon as one in any of your mirrors fails, even if you're not at the machine, it will immediately copy the data onto that drive so you still have redundancy. I did once have a motherboard fail, but I just bought another board with a similar RAID controller on it and it understood the array.

--
A worried father confronted his daughter one night.
"I don't like that new boyfriend, he's rough and common and bloody stupid with it."
"Oh no, Daddy," the daughter replied, "Fred's ever so clever, we've only been going out nine weeks and he's cured me of that illness I used to get once a month."

On Tue, 22 May 2018 21:50:28 +0100, Paul wrote:

Jimmy Wilkinson Knife wrote:
On Tue, 22 May 2018 08:20:46 +0100, Lucifer Morningstar

SSDs are not used in servers due to their unreliability.

They are MORE reliable than rotating rust disks.

I they're not used in servers it's because they can't stand the higher
amount of writing. And that depends on what you're doing with the
server. If i made a server where I wanted fast disk access, but it
wasn't written in huge quantities, I'd use SSDs.

https://ark.intel.com/products/97164...#tab-blade-1-0

# Capacity 750 GB
# Endurance Rating (Lifetime Writes) 41000 TBW
# Mean Time Between Failures (MTBF) 2 million hours
# Enhanced Power Loss Data Protection Yes === "I'm an enterprise product"

That drive can be written end to end, about 55000 times,
if you believe the endurance rating.

Nice if you need that much reliability, but 10 times the price of a standard SSD!
I think I'd prefer to buy 10 times as many normal SSDs for the same money and increase redundancy/speed/capacity.

There are also flash based storage subsystems with internal
RAID, such that if a chip fails, another (unused) chip can
be put in place of it, and rebuilt using redundancy info
in the chip array. I think this is called RAIN.

https://www.micron.com/~/media/docum...f_ssd_rain.pdf

Yes, Crucial SSDs now use this (they're the same company as Micron).

It's possible there are also flash drives (think of a 4U form
factor), where flash chips are arranged on cards, and can
be replaced hot when they fail. Take the top off, plug in a
new card (wait for rebuild). These are products you won't find
on the web, and if you have to ask the price, you can't afford
one. (It's similar to PCI Express, where there are all sorts of
whizzy PCI Express technologies out there, which aren't advertised
on the web. Those might be used in HPC environments.)

Most of the advertising we see, is for the "rubbish stuff" :-)

--
Keith was explaining to his sister how to jump start a car.
"I explained about which cables to hook up where and in what order.
She said, 'ok, I got all that, so now, which car do you start first?'"

Doomsdrzej wrote:
On Mon, 21 May 2018 17:23:41 -0300, pjp
wrote:

In article , lid
says...
On Thu, 17 May 2018 07:49:55 +1000, Peter Jason wrote:

Do these thumb drives last forever, or should
their contents be transferred to the latest USB
drives?
I can only speak from personal experience. I've got several thumb
drives, all from SanDisk.

- 512MB Cruzer Micro
- 4GB Cruzer Titanium
- 16GB Cruzer Contour
- 64GB Extreme (USB 3.0)
- 128GB Extreme Pro (USB 3.0)

I ordered the 64GB and 128GB drives from a shop in Hong Kong (through
eBay).

The 64GB drive is used /all the time/. I transfer MKV files to it, watch
them on my BluRay disc player and then I delete them. I've been doing
that for years now. Never an error. On a rare occasion, the MKV freezes
when playing, but that could also be caused by the file itself or the
disc player. Rebooting the disc player always solves the problem.

I once bought a (supposedly) 4GB flash drive from DaneElec. Never got it
to work properly and it certainly wasn't 4GB.
Bought 5 Duracell branded 64Gb flashdrives. Three of them have gone into
"write protected" mode and can't do anything but read the corrupted data
(some of it) that you can view a listing for.

Duracell has no business producing USB drives.

Duracell is a trademark you can "buy" and slap on stuff.

https://en.wikipedia.org/wiki/Duracell

"Duracell entered into a brand licensing agreement with flash memory
manufacturer Dane-Elec in 2008 for a line of products including
memory cards, hard drives and USB flash drives with the Duracell
brand mark and in the brand's trademark "copper top" coloring."

So Duracell wouldn't know anything about those particular sticks.

And Dane-Elec could be buying Shanghi Surprise sticks and putting
the Duracell name on them. It's not like Duracell gets a say in
whether rubbish has their branding on it.

Just as you can find "General Electric", or "Westinghouse" products,
which neither company knows anything about.

Paul

On Tue, 22 May 2018 16:16:44 -0400, Paul
wrote:

Jimmy Wilkinson Knife wrote:


At what age does the data become unreadable if the drive has not been
powered up?

Flash is generally quoted as holding charge on a floating
gate for around 10 years.

That means, if you're using SSDs for archival storage, they
should be plugged in and re-written every five years, at a guess.

For a given design, I don't know how to guess at that value,
and the 10 year number is merely a "starting point, ball park number".

I'd suggest that every 5 years, you buy a new, faster, AND cheaper
newer flash, and move everything to it. You can consolidate while
you're at it.

"Paul" wrote in message
news
wasbit wrote:
"Paul" wrote in message
news

Time for the testing tools.

https://www.raymond.cc/blog/test-and...-with-h2testw/


######################
Chip Genius

Controller Vendor: Alcor Micro
Controller Part-Number: AU6989SN-GTB/AU6998SN [F206] - F/W FA02
Flash ID code: ADDE14AB - Hynix H27QCG8D2F5R - 1CE/Single Channel
[MLC-16K] - Total Capacity = 8GB

Proving once again, that you really can get $13 worth
of storage, by paying $13 :-)

I take it, when they modified the declaration on the
drive, the controller was limited to declaring 2TB
and could not declare a larger number. Or they would
have set it to an even higher number.


So how does this magic trick work?
The drive has been formatted, rewritten with 1242 (242 GB) videos & several
instances of your big.bin files. Windows shows 411 GB used & 505 GB free
space.
Having transferred the drive from my desktop to a laptop running Linux Mint,
(I'm a novice), so as to isolate it from the original file source, a
notification pops up saying all the contents could not be displayed. Free
space is 543 GB because several of the bin files aren't showing.
All the video files played from a random selection.
To my mind, it would seem there is much more than 8 GB available although I
wouldn't trust the drive for anything important.

--
Regards
wasbit

wasbit wrote:
"Paul" wrote in message
news
wasbit wrote:
"Paul" wrote in message
news
Time for the testing tools.

https://www.raymond.cc/blog/test-and...-with-h2testw/



######################
Chip Genius

Controller Vendor: Alcor Micro
Controller Part-Number: AU6989SN-GTB/AU6998SN [F206] - F/W FA02
Flash ID code: ADDE14AB - Hynix H27QCG8D2F5R - 1CE/Single
Channel [MLC-16K] - Total Capacity = 8GB

Proving once again, that you really can get $13 worth
of storage, by paying $13 :-)

I take it, when they modified the declaration on the
drive, the controller was limited to declaring 2TB
and could not declare a larger number. Or they would
have set it to an even higher number.


So how does this magic trick work?
The drive has been formatted, rewritten with 1242 (242 GB) videos &
several instances of your big.bin files. Windows shows 411 GB used & 505
GB free space.
Having transferred the drive from my desktop to a laptop running Linux
Mint, (I'm a novice), so as to isolate it from the original file source,
a notification pops up saying all the contents could not be displayed.
Free space is 543 GB because several of the bin files aren't showing.
All the video files played from a random selection.
To my mind, it would seem there is much more than 8 GB available
although I wouldn't trust the drive for anything important.


You can disassemble the Flash stick and verify the chip
part number, if you don't believe the output of the
Chip Genius. That's one way to resolve any uncertainty.

*******

The controller can have a couple of behaviors.

1) Chip designers never intended to support fraud.
2) Chip designers actively supported the sale of fraudulent sticks.

In the first one, there is a single register, and by
inflating the capacity to 2TB, the address "rolls over"
when writes pass the 8GB point. Say, for example, you
wrote 242GB of material to the stick. The address
used at that point would be (242GB mod 8GB) and the
stick would actually be writing blocks at the 2GB point.
242 divided by 8 goes 30 times, with 2 left over as
a remainder. And that's what the "mod" tells you.

If the design worked that way, then Windows should be
quite upset, right after you write slightly more than
8GB of materials. If you wrote to 8GB+one_sector,
then the NTFS file header would get overwritten
by the file you were trying to transfer. Windows
should be ****ed.

The other behavior they could put in the design, is two
registers. One register contains the "real" chip capacity
of 8GB, while a second register is used as a value returned
when Windows queries the available device size. By doing
it this way, instead of implementing address rollover with
the mod() function, they can instead "clip" the address.
Then, any address greater than 8GB, simply results in the
write operation being blocked. This ensures integrity for
anything written below 8GB, but nothing written above 8GB
comes back with anything other than zeros. This behavior
requires the designers to be "in on the fraud concept"
and using two registers for the express purpose of allowing
"inflated" sticks to be sold.

But no matter how you slice it, if the chip really is 8GB,
it cannot hold more than 8GB of data. All it takes is
someone making a test file from /dev/random, a file with
high entropy, that will defeat any compression mechanisms
and ensure an honest test result. And they'll immediately
be able to tell it isn't a 2TB drive.

Paul

On Wed, 23 May 2018 06:24:51 -0400, Paul
wrote:

wasbit wrote:
"Paul" wrote in message
news
wasbit wrote:
"Paul" wrote in message
news
Time for the testing tools.

https://www.raymond.cc/blog/test-and...-with-h2testw/



######################
Chip Genius

Controller Vendor: Alcor Micro
Controller Part-Number: AU6989SN-GTB/AU6998SN [F206] - F/W FA02
Flash ID code: ADDE14AB - Hynix H27QCG8D2F5R - 1CE/Single
Channel [MLC-16K] - Total Capacity = 8GB

Proving once again, that you really can get $13 worth
of storage, by paying $13 :-)

I take it, when they modified the declaration on the
drive, the controller was limited to declaring 2TB
and could not declare a larger number. Or they would
have set it to an even higher number.


So how does this magic trick work?
The drive has been formatted, rewritten with 1242 (242 GB) videos &
several instances of your big.bin files. Windows shows 411 GB used & 505
GB free space.
Having transferred the drive from my desktop to a laptop running Linux
Mint, (I'm a novice), so as to isolate it from the original file source,
a notification pops up saying all the contents could not be displayed.
Free space is 543 GB because several of the bin files aren't showing.
All the video files played from a random selection.
To my mind, it would seem there is much more than 8 GB available
although I wouldn't trust the drive for anything important.


You can disassemble the Flash stick and verify the chip
part number, if you don't believe the output of the
Chip Genius. That's one way to resolve any uncertainty.

*******

The controller can have a couple of behaviors.

1) Chip designers never intended to support fraud.
2) Chip designers actively supported the sale of fraudulent sticks.

In the first one, there is a single register, and by
inflating the capacity to 2TB, the address "rolls over"
when writes pass the 8GB point. Say, for example, you
wrote 242GB of material to the stick. The address
used at that point would be (242GB mod 8GB) and the
stick would actually be writing blocks at the 2GB point.
242 divided by 8 goes 30 times, with 2 left over as
a remainder. And that's what the "mod" tells you.

If the design worked that way, then Windows should be
quite upset, right after you write slightly more than
8GB of materials. If you wrote to 8GB+one_sector,
then the NTFS file header would get overwritten
by the file you were trying to transfer. Windows
should be ****ed.

The other behavior they could put in the design, is two
registers. One register contains the "real" chip capacity
of 8GB, while a second register is used as a value returned
when Windows queries the available device size. By doing
it this way, instead of implementing address rollover with
the mod() function, they can instead "clip" the address.
Then, any address greater than 8GB, simply results in the
write operation being blocked. This ensures integrity for
anything written below 8GB, but nothing written above 8GB
comes back with anything other than zeros. This behavior
requires the designers to be "in on the fraud concept"
and using two registers for the express purpose of allowing
"inflated" sticks to be sold.

But no matter how you slice it, if the chip really is 8GB,
it cannot hold more than 8GB of data. All it takes is
someone making a test file from /dev/random, a file with
high entropy, that will defeat any compression mechanisms
and ensure an honest test result. And they'll immediately
be able to tell it isn't a 2TB drive.

I got caught in a scam where I bought what I thought was
a 4 gig mp3 player. Turned out to be a 1 gig player that
had been hacked to show 4 gig to windows.
Once I tried to put more than 1 gig on it windows complained
of write errors.

Paul

On Wed, 23 May 2018 10:16:19 +0100, "wasbit"
wrote:

"Paul" wrote in message
news
wasbit wrote:
"Paul" wrote in message
news
Time for the testing tools.

https://www.raymond.cc/blog/test-and...-with-h2testw/


######################
Chip Genius

Controller Vendor: Alcor Micro
Controller Part-Number: AU6989SN-GTB/AU6998SN [F206] - F/W FA02
Flash ID code: ADDE14AB - Hynix H27QCG8D2F5R - 1CE/Single Channel
[MLC-16K] - Total Capacity = 8GB

Proving once again, that you really can get $13 worth
of storage, by paying $13 :-)

I take it, when they modified the declaration on the
drive, the controller was limited to declaring 2TB
and could not declare a larger number. Or they would
have set it to an even higher number.


So how does this magic trick work?
The drive has been formatted, rewritten with 1242 (242 GB) videos & several
instances of your big.bin files. Windows shows 411 GB used & 505 GB free
space.
Having transferred the drive from my desktop to a laptop running Linux Mint,
(I'm a novice), so as to isolate it from the original file source, a
notification pops up saying all the contents could not be displayed.

If that message was correct none of the files could be displayed
yet you say some still showed.

Free
space is 543 GB because several of the bin files aren't showing.
All the video files played from a random selection.

So the message was some of the contents could not be displayed.

To my mind, it would seem there is much more than 8 GB available although I
wouldn't trust the drive for anything important.

Google the device number or add more data until it gives write errors.