Tape Backup

[J. Clarke]

There is a PCMCIA SATA adapter. It's possible I read it in a review in
Tom'shardware in a review of SATA external exclosures.

Actually there are several available. Newegg carries three, and froogle has
70 hits including at least one brand that newegg doesn't carry.

I have a Buslink, which is basically just a Silicon Image chip on a board in
a case and seems to run on the Silicon Image reference drivers. So far it
works fine.

They're quite remarkably cheap.

 

[J. Clarke]

I used it to copy 5 GB from the external drive to the laptop - worked
fine.

But how long did it take compared to doing it to an SATA device? And if the
performance was good, what bridge chip was it using?

 

[J. Clarke]

Lots of MAC users run heavy applications (mostly Photoshop) right
of a fireware external device. Performance is OK.

Photoshop wouldn't seem to me to be a "heavy application" from a disk I/O
viewpoint.

 

[Mike Tomlinson]

Of course, figuring out
what was supposed to be on there and whether it is still good or not
won't be easy!

CD-Check is a very useful utility. google for it.

 

[Neil Maxwell]

The thin layer of laquer on the label side can be worn off over
time (through mishandling, stacking CDs on top of each other etc),
causing "CD Rot" as the aluminium surface oxides.
Even the pressure of a pen is enough....

This is true, and is important to be aware of, though good quality
disks have a protective top layer that is more resistant to pressure
damage than cheap ones. This is easily tested on junk CDRs of
different types and quality.

The real problem is chemical in nature, I believe, and is due to the
dye changes relaxing over time. I've seen disks stored in a jewel
case since they were written, in a cool, dark location, that have
failed to read 2 years later. Definitely not abrasion/mishandling,
UV, or other more quantifiable bugaboos.

I believe the CDR industry knows exactly what's going on, but they're
not interested in highlighting it as an issue. It'll all break out
one day, and it'll be an interesting mess when it does.

 

[Neil Maxwell]

Gee, this is terrible. According to the NIST thing, I am doing so many
things wrong. I write on them with a broad permanent marker which sure
smells like it uses solvents, sometimes with fine tip markers, and I
store the recorded media horizontally for years, sometimes in their own
cases laying down piled on top of each other, but usually just back in
the spindles they came in.

Note that NIST has virtually no hard data that they refer to for
support. There are not comprehensive in-depth studies of CDR data
degradation that I know of beyond simple accelerated failure testing,
which is clearly insufficient.

 

[Neil Maxwell]

High quality CDR (e.g. Mitsui Archive Gold) have undergone a lot of
testing and seem to be quite stable for long periods. The jury is
still out for DVDR. Hard drives contain all kinds of seals, filters,
lubricants on mechanical parts, and flash memory parameters and
firmware dependent on floating charges, all of which can decay over a
period of years. Hard drives are quite unreliable for long term
storage.

Sorry, I've got a bunch of Mitsui and Kodak Gold and Silver from a few
years back that are nearly dead. I can't read them on most of my
readers and burners (multiple Plextors, NECs, Teacs, no-names), but my
Lite-on LTD163 will pull data that other drives give up on.

The only reliably readable disks I have from the past 6-7 years are
TY, and I've only been using them extensively for a couple of years,
so the data's not solid on them yet.

YMMV, as always.

 

[J. Clarke]

This is true, and is important to be aware of, though good quality
disks have a protective top layer that is more resistant to pressure
damage than cheap ones. This is easily tested on junk CDRs of
different types and quality.

The real problem is chemical in nature, I believe, and is due to the
dye changes relaxing over time. I've seen disks stored in a jewel
case since they were written, in a cool, dark location, that have
failed to read 2 years later. Definitely not abrasion/mishandling,
UV, or other more quantifiable bugaboos.

I believe the CDR industry knows exactly what's going on, but they're
not interested in highlighting it as an issue. It'll all break out
one day, and it'll be an interesting mess when it does.

The dye matters--some are more stable than others. This is well known and
is not news to anyone. How stable depends on who you talk to.

CD rot was a problem that occurred with prerecorded CDs, and has been
corrected with manufacturing changes I understand.

I'm not sure I understand what you think the CDR industry "knows is going
on" or why it would be "an interesting mess" when it "all breaks out" one
day.

 

[Al Dykes]

Photoshop wouldn't seem to me to be a "heavy application" from a disk I/O
viewpoint.

Photoshop benefits greatly when you add a dedicated "work disk", or a
dedicated partition on a second disk. The faster your computer the
more likely your disk is to be the bottleneck and PS has settings for
up to 3 work disks. A work disk makes my machine faster, and I've got
1 GB ram.

I know a PS instructor who carries a MAC with an external FW disk that
has PS on it. She gets performance good enough to use in front of a
class, but in class she avoids using many layers (a PS term) that
multiply the size of the image and slow things down. People that use
PS to the fullest can easily wind up with a 100MB PSD file and keep it
in that format, for future rework.

 

[Neil Maxwell]

The dye matters--some are more stable than others. This is well known and
is not news to anyone. How stable depends on who you talk to.

It is news to some, and most people are unaware of the issue, IME.
Lots of people believe the marketing claims of multi-decade life for
CDRs, or believe that the main culprit in failure is physical damage,
as demonstrated in this very thread.

CD rot was a problem that occurred with prerecorded CDs, and has been
corrected with manufacturing changes I understand.

Yep, different issue, brought up by someone else.

I'm not sure I understand what you think the CDR industry "knows is going
on" or why it would be "an interesting mess" when it "all breaks out" one
day.

The CDR industry, in general, continues to make claims regarding media
and data lives that I regard as unsupportable. I have no doubt that
they are very aware of the dye-related lifetime issues, and I also am
unaware of this being highlighted anywhere by CDR manufacturers. Note
that I have no links or data to support my assumption of their
awareness of the issues, but I'm pretty confident about it.

As a result, there's a common misconception that burned CDRs have very
long lifetimes, and that data burned to CDRs is safe. People who are
aware that there are problems, and that it's a relatively complex
issue, are in a small minority.

Eventually, awareness may reach a critical mass where users in general
know that this is a concern. If and when this happens, it may become
a cause celebre that generates much heat and furor. Or not. I'd like
to see it happen, but I'm fond of accurate, data-based marketing,
which is admittedly idealistic and even unrealistic.

I'm not sure if you're just picking a fight here; none of this is
rocket science if you're aware of the issues, as you seem to be.

 

[J. Clarke]

Photoshop benefits greatly when you add a dedicated "work disk", or a
dedicated partition on a second disk. The faster your computer the
more likely your disk is to be the bottleneck and PS has settings for
up to 3 work disks. A work disk makes my machine faster, and I've got
1 GB ram.

I know a PS instructor who carries a MAC with an external FW disk that
has PS on it. She gets performance good enough to use in front of a
class, but in class she avoids using many layers (a PS term) that
multiply the size of the image and slow things down. People that use
PS to the fullest can easily wind up with a 100MB PSD file and keep it
in that format, for future rework.

How does that same disk work for analog video capture?

 

[Sayso Takewashi]

Gee, this is terrible. According to the NIST thing, I am doing so many
things wrong. I write on them with a broad permanent marker which sure
smells like it uses solvents, sometimes with fine tip markers, and I
store the recorded media horizontally for years, sometimes in their own
cases laying down piled on top of each other, but usually just back in
the spindles they came in.

Funny thing, though, I have never, ever noticed a disc lose data it
once held. Could it all just be a lot of hype? Wouldn't be the first
time. So now I have to go back and try a few. Of course, figuring out
what was supposed to be on there and whether it is still good or not
won't be easy!

Wish me luck,

On Weekend i recorded 10 CD-Rs,two of them have Read-Errors 2 Hours
after recording.I handled them carefully,put them in a CD-Map and
would only grap them to read the contents...

I would not trust them,like you do.

 

[Mike Redrobe]

On Weekend i recorded 10 CD-Rs,two of them have Read-Errors 2 Hours
after recording.I handled them carefully,put them in a CD-Map and
would only grap them to read the contents...

Something is very wrong with your burner, or disks.
I`ll agree there may be problems a few years ahead with CDR, but
read-errors after 2 hours shows something is wrong with your setup,
not CDR media as a whole

I would not trust them,like you do.

Come back when you are buying verbatim disks, or have a new burner...

 

[Paul Rubin]

But will a virus typically attack any file that is not executable? While
some do, is that the normal action?

The virus might simply try to reformat a hard disk, or scribble on it
at random places. With an uncompressed disk there's some chance of
recovery; with compressed, encrypted disk, it's more likely hopeless.

It's actually quite easily done. You pull the board off of your off-site
drive and put it on the dead drive, which generally involves about ten
minutes with a screwdriver.

That's the easy part. The hard part is getting your data back after
you do the board swap. All I can say is, the attempts I've personally
had any contact with have been both expensive and ultimately
unsuccessful. But that's not all that many.

Easily done--SATA PCCard adapters go for about 20 bucks.

Then you have to go through the PC card interface which is slower than
a disk interface, if my experience with an Adaptec 1480 scsi adapter
is any indication.

Read what I wrote again. I didn't say "use RAID for backup". I said backup
_to_ the RAID. The step I assumed was obvious was to "then pull both
drives, replace them with the next day's backup set, take one home, leave
the other in the safe".

There is software that does this with tapes. You need two tape drives and
the last time I checked the price the software was a thousand dollar add-in
to a several thousand dollar enterprise backup package.

The equivalent with tape would be to just write two backups, or even
write one backup to two tape drives. That doesn't sound like it needs
a multi-kilobuck software package.

There's a crossover point on very large systems where a tape library
becomes cost effective. For home use the cost of reliable tape is
prohibitive.

Well, if you really want to use two tape drives, that doubles the
cost, but you can get an LTO1 drive for about $600 now, so two of them
cost about what I paid for my home DDS2 drive in the mid 90's. I'd
say the cost is steep for a typical home user but I'd stop short of
"prohibitive".

Are you using just one disk or are you using a set of them in a rotation
backup like you would with tapes?

I'm using them more like archival tapes, i.e. write-once, no rotation.

I have a number of tape drives. The trouble with them is that disk
capacity is increasing faster than tape capacity, and you need a
state of the art tape to back up a cheap disk.

That describes the situation really well. Right now, disk capacity
increase seems to have stalled, while tape is making some significant
advances, so for the first time in a while, high-end tape cartridges
hold more than a high-end disk drive's worth of data. I don't know if
that will last, and those tape systems really are too expensive for
home use (LTO3, SAIT).

And quite honestly, I'd trust disk over DDS. I've had DDS drives eat
multiple tapes.

I think a DDS backup might be more likely to fail at the moment that
it's made and there's always a chance of a drive eating a tape. I
haven't (yet) had bad experiences with tapes going south while sitting
on a shelf, which I've had with disk drives more than once.

On my infinite to-do list is to write some backup software that
uses a RAID-like strategy so if your backup needs, say, 5 tapes,
you can instead write it on (say) 8 tapes, and then if any three
of them get trashed you can still restore from the remaining 5.

 

[Al Dykes]

Something is very wrong with your burner, or disks.
I`ll agree there may be problems a few years ahead with CDR, but
read-errors after 2 hours shows something is wrong with your setup,
not CDR media as a whole


Come back when you are buying verbatim disks, or have a new burner...

Did you do a readback/verify when you made them. Were that OK then ?

 

[J. Clarke]

The virus might simply try to reformat a hard disk, or scribble on it
at random places. With an uncompressed disk there's some chance of
recovery; with compressed, encrypted disk, it's more likely hopeless.

Compressed, encrypted file on a disk. Not a compressed, encrypted disk.

That's the easy part. The hard part is getting your data back after
you do the board swap. All I can say is, the attempts I've personally
had any contact with have been both expensive and ultimately
unsuccessful. But that's not all that many.


Then you have to go through the PC card interface which is slower than
a disk interface, if my experience with an Adaptec 1480 scsi adapter
is any indication.

Cardbus is 32 bit 33MHz, just like PCI. And even if it is slower than the
disk interface, no disk on the market can fill even a 100 MB/sec pipe.

The equivalent with tape would be to just write two backups, or even
write one backup to two tape drives. That doesn't sound like it needs
a multi-kilobuck software package.

If you write two backups then it takes twice as long and since tape is slow
anyway . . .

Now, actually backup to two tapes simultaneously and tell me how you did it
that didn't cost kilobucks.

Well, if you really want to use two tape drives, that doubles the
cost, but you can get an LTO1 drive for about $600 now, so two of them
cost about what I paid for my home DDS2 drive in the mid 90's. I'd
say the cost is steep for a typical home user but I'd stop short of
"prohibitive".

Where can you get an LTO1 drive for $600 with any kind of reasonable
expectation that it's not busted or stolen?

And no, I don't want to use two tape drives, I want to have backup. I'm not
wedded to one technology like you seem to be. But you've pointed out that
if you do want to have redundant backup with tape you pay twice for the
drives or take twice as long for the backup.

I'm using them more like archival tapes, i.e. write-once, no rotation.

That's not backup, that's archiving.

That describes the situation really well. Right now, disk capacity
increase seems to have stalled,

It has? 400 gig disks just started shipping for 400 bucks. Where can I get
a 400 gig tape for 400 bucks?

while tape is making some significant
advances, so for the first time in a while, high-end tape cartridges
hold more than a high-end disk drive's worth of data.

They may hold more than a "high-end disk drive" if by that you mean an
enterprise SCSI drive, but they don't hold more than a big consumer disk
drive.

I don't know if
that will last, and those tape systems really are too expensive for
home use (LTO3, SAIT).


I think a DDS backup might be more likely to fail at the moment that
it's made and there's always a chance of a drive eating a tape. I
haven't (yet) had bad experiences with tapes going south while sitting
on a shelf, which I've had with disk drives more than once.

On my infinite to-do list is to write some backup software that
uses a RAID-like strategy so if your backup needs, say, 5 tapes,
you can instead write it on (say) 8 tapes, and then if any three
of them get trashed you can still restore from the remaining 5.

That's commercially available from several vendors.

 

[Paul Rubin]

Compressed, encrypted file on a disk. Not a compressed, encrypted disk.

OK, that too. Actually maybe even worse, depending on the type of
encryption and authentication. Change one bit and the whole file is
potentially turned into garbage.

Cardbus is 32 bit 33MHz, just like PCI. And even if it is slower than the
disk interface, no disk on the market can fill even a 100 MB/sec pipe.

Yes, however, I've gotten crap performance through that Cardbus 1480
card. Maybe something else was wrong that I didn't try to pursue.
Have you actually measured performance of external discs through PC
card SATA?

If you write two backups then it takes twice as long and since tape is slow
anyway . . .

Eh? The problem with tape that I'm facing is that it's too fast, not
too slow. LTO1 has around 15MB native transfer speed which a PC disk
system -might- be able to keep up with. LTO2 is around 30MB/sec which
is hard to keep fed. LTO3 ($$$) is almost 70MB/sec. I just don't see
how to back up to LTO3 at full speed from any type of PC.
(Figures from http://www.span.com/catalog/index.php?cPath=21_45).

Now, actually backup to two tapes simultaneously and tell me how you
did it that didn't cost kilobucks.

I think the bottleneck would be hardware transfer speeds, not software.

Where can you get an LTO1 drive for $600 with any kind of reasonable
expectation that it's not busted or stolen?

There were several new units on ebay for around that price last week.
LTO1/LTO2 have gotten cheaper since LTO3 started shipping recently.

It has? 400 gig disks just started shipping for 400 bucks. Where can I get
a 400 gig tape for 400 bucks?

http://www.tape4backup.com/newlookltobrand.html
Scroll down to the LTO3 selection.

They may hold more than a "high-end disk drive" if by that you mean an
enterprise SCSI drive, but they don't hold more than a big consumer disk
drive.

LTO3 holds 400GB and SAIT-1 holds 500GB. SAIT-2 (scheduled for next
year) is supposed to hold 1000GB and there are no disk drives with
that capacity expected by then.

That's commercially available from several vendors.

I only use free software, but it's possible that there's some already
available.

 

[Malcolm Weir]

On Tue, 18 Jan 2005 18:00:42 -0500, "J. Clarke"

[ Snip ]

If you write two backups then it takes twice as long and since tape is slow
anyway . . .

Nonsense.

Tape is not slow. Your basic LTO1 will do at least 7.5MB/sec
uncompressed, LTO2 will do 15MB/sec uncompressed, and LTO3 will do
30MB/sec uncompressed. And they will all compress, too.

Feeding those things (and similar spec devices) is *hard*.

Because *disks* are slow... for these sorts of huge sequential
operations!


Malc.

 

[J. Clarke]

OK, that too. Actually maybe even worse, depending on the type of
encryption and authentication. Change one bit and the whole file is
potentially turned into garbage.


Yes, however, I've gotten crap performance through that Cardbus 1480
card. Maybe something else was wrong that I didn't try to pursue.
Have you actually measured performance of external discs through PC
card SATA?

Haven't measured, but was able to copy 35 gig in about a half an hour using
LiveState Recovery.

Eh? The problem with tape that I'm facing is that it's too fast, not
too slow. LTO1 has around 15MB native transfer speed which a PC disk
system -might- be able to keep up with. LTO2 is around 30MB/sec which
is hard to keep fed. LTO3 ($$$) is almost 70MB/sec. I just don't see
how to back up to LTO3 at full speed from any type of PC.
RAID.

(Figures from http://www.span.com/catalog/index.php?cPath=21_45).

Believe it when you see it. The performance of _anything_ is inflated in
the advertising.

I think the bottleneck would be hardware transfer speeds, not software.

Regardless of where the bottleneck is you need to have software which will
write the same data to two tapes simultaneously in order to backup to two
drives simultaneously.

There were several new units on ebay for around that price last week.

Is that the price they finally went for or the starting price? Lowest ebay
price I've seen is a bit less than the lowest Froogle price but that
auction still had some time to go. And they're still in the $800 range.

LTO1/LTO2 have gotten cheaper since LTO3 started shipping recently.


http://www.tape4backup.com/newlookltobrand.html
Scroll down to the LTO3 selection.

And what am I supposed to see there? I'm sorry, but a $130 cartridge and a
$5000 tape drive is not 400 gig for 400 bucks, it's 400 gig for 5130 bucks.

LTO3 holds 400GB and SAIT-1 holds 500GB. SAIT-2 (scheduled for next
year) is supposed to hold 1000GB and there are no disk drives with
that capacity expected by then.

What's expected and what happens are two different things. Regardless,
that's a $6K drive and if the only vendor is Sony then I'd rather trust my
data to a shredder thank you.

I only use free software, but it's possible that there's some already
available.

So far you have convinced me of nothing. Perhaps if you weren't buying
$6000 tape drives you could afford some software.

 

[J. Clarke]

On Tue, 18 Jan 2005 18:00:42 -0500, "J. Clarke"

[ Snip ]

If you write two backups then it takes twice as long and since tape is
slow anyway . . .

Nonsense.

Tape is not slow. Your basic LTO1 will do at least 7.5MB/sec
uncompressed, LTO2 will do 15MB/sec uncompressed, and LTO3 will do
30MB/sec uncompressed. And they will all compress, too.

Feeding those things (and similar spec devices) is *hard*.

Because *disks* are slow... for these sorts of huge sequential
operations!

So we spend $12,000 for hardware to back up a $200 disk? What's wrong with
this picture?