How To Verify CD-R Data Retention Over Time?

Peter (Professor) Fo writes "I've recently had two CD-Rs reported to me as faulty which are just 3 years old. This is worrying — I suspect the failure rate for this batch could be 10%. When researching CD longevity there is old and unreliable information; pious 'how to cosset your discs so they last 100 years' blurb; and endless discussions of what sort of dye to use, don't use cheap media, burn slower (or don't), but not much by way of hard facts besides there's a lot of data loss going on. Does anyone know of a generic utility (win or *nix would suit me) that can map sector readability/error rates of CDs? I'd like to measure decay over time in my environment with my media and my other variables; and I expect others would too."

Comments

dvdisaster

[Anonymous Coward]

You should probably try dvdisaster. it can test media, and can create (on disk or external) redundancy data, which can be used to recover later.
It's also open source, so you could probably coerce it to export some more information

Re:dvdisaster

[dougmc]

Very nice. I use par2 for basically the same purpose. I save about 10% of my DVD capacity, and have a program that creates a directory with md5sums of every file, along with par2 files for all the files, so I can recover from a loss of almost up to 10% of the disk's data.

Of course, if the data lost is in the catalog so I can't even find my files, then things get much more complicated. But even so, I've had to use this system a few times (due to damaged DVDs mostly) and it's worked pretty well.

Re:dvdisaster

[Reziac]

Not true of DVDs, which to my understanding can deteriorate even without damage.

I've bought new-in-box DVDs that were defective, without any visible damage. :(

Re:dvdisaster

[hansamurai]

10 cents is pushing it, I'd say more around 13 cents per gigabyte, but DVD media sits happily at 5-6 cents a gigabyte. Of course there are obvious differences, but my mantra is "burn and forget." I don't need to be bothered taking up hard drive space for a movie or iso I may need once a year. I just file the disc away in a large stack and fish it out if I ever need it.

Though as hard drives continue to fall in price and media stagnates (at least this generation), I will probably eventually buy a very large one and just use that. Of course if that one hard drive dies I lose a lot, but if one disc gets scratched, I only lose so much. RAID or backups only add to the per gigabyte cost.

Re:dvdisaster

[Reziac]

Yeah, that's common enough... but even pressed DVDs are, to my understanding, inherently more fragile than they appear, and highly subject to environmental degradation. Any unsealed microhole and eventually the data layer is toast. :(

As to the inverse bell curve, I see it regularly in OEM computers (any brand) -- fairly high DOA rate, and after that... IMO they're designed to fail** due to always being on the edge of overheated. Those that survive past the designed-in time of death (3 to 5 years) usually live forever.

** Don't think so? stock top-of-the-line Dell, ran hot enough that it tended to lock up... I gave it the same very basic cooling any cheap clone gets, and its internal temp dropped FORTY degrees.

par2

[Drinking Bleach]

Whenever I back up important data, I use par2. If the disc has I/O errors, I just make a full image with dd_rescue (skips past bad blocks, whereas dd will just halt operation) and run "par2 verify" on it. If it's really important, I always verify the integrity no matter what (I've even done it on discs 2 days old, and sometimes, due to the reliability of CD/DVD-R media, it even has errors to repair).

Re:par2

[zippthorne]

Problem with par2 is that it does not work well with lots of files spread out across directories, and it especially doesn't work recursively. Though par2-ing your image is still probably a pretty good idea.

My partial solution so far is to make a hidden directory and hard link *every* file on the disk into that directory, then run par2 in that "flat" directory. (I put the inode number in the filenames to make sure there aren't any name collisions. If you use the inode number AS the name, you won't get any dupes on multiply linked files)

I haven't got the details worked out yet, though, but iso9660 supports hard links, so the disk burns and verifies well, and (importantly to me) unlike dvdisaster, it's not just a technically usable image, but an actually standard-compliant image.

I'm not quite sure how I'll go about restoring if verify fails, though, since the goal is to keep the same directory structure (presuming it's still mostly intact) and repair the files themselves. Perhaps unionfs over the loopback mounted iso would work. Or just repairing the files and using the a recursive directory listing (also saved on disk) to regenerate the directory structure.

But this would all be less necessary if the standard ECC in the CD format was more generalized: i.e. you could set it to always use the entire disk, no matter how little actual data you had, and just use extra layers of ECC that take advantage of the extra space. It would be good if reading could still be done transparently (like current CD ECC) and report "goodness" of the disk as determined by how much of the parity data was actually needed (so you could keep track of degradation over time, if any, and replace disks if the numbers start increasing)

Re:par2

[Anonymous Coward]

PAR 1.x worked as you describe, on a per-file basis.

PAR 2.x divides all data across multiple files into blocks. If 75 KB worth of data is corrupt on a group of files including a 2 GB file, a 1 MB file, and a 5 KB file, you only need about 75 KB worth of PAR2 files to correct everything. Of course it is extremely slow to break everything into very small block sizes and generate the appropriate PAR2 files, so you'll probably have 1 MB or 5 MB blocks (or whatever) to deal with instead. In the example, if that 75 KB worth of bad data across three files lies in four separate blocks, and you have 5 MB blocks, it will take 20 MB worth of PAR2s to fix the problem. This dedication of storage space is still favorable to retaining multiple redundant copies of the original data, as long as you can accept the trade-off of time spent in the generation of the parity files.

Error-Proofing Data With Reed-Solomon Codes

[MarkoNo5]

The following tool allows you to track the failure rate of your media, and allows you to recover the files and replace the cd/dvd when it starts failing. http://hardware.slashdot.org/article.pl?sid=08/08/03/197254

Save them again with long lasting solutions

[Maxwell42]

If you consider your data worth it, have enough time and enough money, you should probably re-burn/re-save them to long lasting media.

There was a previous post on askslashdot about this subject.
http://ask.slashdot.org/article.pl?sid=08/08/27/2119252

My suggestion was to use Plasmon "Century-Disc" :
http://slashdot.org/comments.pl?sid=914095&cid=24784787
(even though I have never tried it myself)

Add PAR2 files

[Still an AC]

When archiving data to CD or DVD I use a program call Quick Par to generate recovery files that I can use to repair the data on the disc if it becomes damaged.
It is based on the same recovery tech that RAID systems w/ parity drives use, and is mostly used to repair Usenet downloads. I usually put 4GB of data and 400MB of PAR2 recovery files on the disc. This will allow ~10% data loss before recovery is not possible. Also I dont have to worry about the TrueCrypt vol becoming damaged and unusable as well.

Try the CDFreaks Forums

[Wanker]

The obsessed people at CDFreaks can help. Here's a link to their FAQ on CD-R media:

http://club.cdfreaks.com/f33/media-faq-61943/

In other places in the cdfreaks forums, you'll find links to tools that can read the C1/C2 error rates. One of the simplest is "readcd", part of the "cdrecord" programs on Linux.

In the DVD world, Lite-On and Plextor both make proprietary programs to read the media-level error rates which only work with their own drives. Lite-On has a Linux version of theirs.

Re:Try the CDFreaks Forums

[Opyros]

Note that on Debian-derived distros, the equivalent of "readcd" is called "readom".

Professional Advice

[polyomninym]

I work with CD/DVD and related technology as a profession. I analyze, QC, and mass duplicate media by the thousands for extreme and critical field use, every day. My best advice to you is to use Taiyo-Yuden (TY) media, always. I've seen mixed results and bad burns from Mitsui and MAM-A gold, Kodak, and the like. The TY dye type has a proven longer longevity than any other so far. I also suggest burning all of your audio CD's at 16X, this affect what's called single-beam readers. Also, it insures higher integrity of the burn. Burn 16X DVD's at 8X to increase the write integrity.

What others say about is CDCheck is true, use it along with this advice. Use Plextools Pro on a PX-716 drive if you can find one. It seems to be more accurate than Nero tools. Use Plextools to check the C1, C2, and CU rates. If the graph is half-way to the top of the reading, back that disc up. As cheap as media is, I suggest burning more than one copy, storing the image on an external archive hardrive. When burning, don't use overburning. You lose some integrity for error correction.

Store your media in a cool dry place, on it's side. Avoid humidity, light, and heat when you can. Remember, the best analysis tools in the industry are very expensive for individuals. Take a look at CATs if you are interested in learning more about optical media testing. Best wishes!

use CD/DVD speed

[YesIAmAScript]

http://www.cdspeed2000.com/

You need certain brands of optical drives, but with them and this program (and others), you can see the PI/PO or C1/C2 correction (I can't remember which is for CD and which is for DVD) rates on a per-sector basis on your disc. As the rates rise, the disc is going bad, becoming marginally readable and you can copy the disc before it becomes unreadable.

You can find out which drives to buy at http://cdfreaks.com/. The terminology on there for a drive that can do this is a "scanning drive".

I have no idea if you will find that your correction rates are rising over time.

Re:use CD/DVD speed

[sricetx]

FYI - Nero CD Speed version 4.7.7.16 runs fine in Linux under recent versions of Wine. The newer version (callled Disc Speed, not CD Speed) does not work under Wine.

I've used it with two different Liteon SATA drives: a Liteon 20A3S and a Liteon 20A1L. Both of these drives (and I believe, Plextors) support scanning for jitter. When you run CDSpeed, the test you want is the Disc Quality tab. Click Advanced and then check the DVD Jitter checkbox. This test will give a good an indication of the quality of the disc.

In my experience, Verbatim Datalife Plus (media code: MCC) are really good discs. CMC Magnetics media vary widely, some are okay, others are garbage.

Re:Ask people in the music industry...

[Drinking Bleach]

CDs aren't burned for commercial distribution, they're pressed.

Re:It's ok...

[hjf]

http://dar.linux.free.fr/ try dar. It's like tar but for disks. it also generates PAR files (FEC data) which can help rebuild damaged media.

Re:The tools are called

[digitalchinky]

Short of an electron scanning microscope, the only way to do it is to hook an oscilloscope directly on to the test points within the drive itself and measure signal levels. This will allow you to measure one or both of: Degradation of the laser optics, degradation of the media. It's anyones guess as to which is which :-)

To make things a little more accurate, you should use several drives to test the media. The drives could benefit from being locked away until such a time as they are needed to repeat the tests. Mix in a few new drives when you do actually make your tests in future as well.

Re:Ask people in the music industry...

[Rob Riggs]

Commercial CDs are not burned. They are stamped.

Re:But where to buy TY?

[AxemRed]

You can buy Taiyo-Yuden media at...

http://www.supermediastore.com/

That's where I got mine. I haven't had a bad burn yet.

Re:I highly recommend using Archival Grade Media

[Penguinoflight]

Taiyo Yuden is excellent media, as is most Verbatim media.

To answer a sibling question: no, Taiyo Yuden doesn't make dual layer discs. Verbatim does, but they aren't up to the quality of their single-layer discs.

Re:Mod parent up!

[painandgreed]

3. An external back up HDD or three will fit in a fairly small safety deposit box, so you can recover your data if your house burnsdown or is robbed. A similar number of CDs or even DVDs would require a much larger box to do so and is easier to manage.

Re:If drive makers were SMART, they would...

[ShieldW0lf]

The Exact Audio Copy tool does something like this. (Or, it used to... I kicked the Windows habit a while ago, so I'm not really up to speed)

It only works with Audio CDs to my knowledge, but it will read each section of the disk twice, and if they are identical, it will commit that and move on. If they're not identical, it will read 8 times and look for 4 matches that are identical, and if it finds them, it will commit that. Otherwise, it will read 8 more times and try again, until it's successfully found a match or until it's tried 80 times, at which point it will log the section as corrupt and move to the next sector.

I don't see why it shouldn't be possible to use a similar technique for data disks.

Re:I highly recommend using Archival Grade Media

[Skye16]

I'm not an expert by any stretch, but burnable CDs are all made by using a laser on a reactive film of dye. If the laser sits on a certain spot, it changes the dye a different color. Thus, it imitates the pits from a normally pressed CD/DVD.

The problem is that over time, this dye begins to break down. Your pits stop looking like pits, or non-pits start looking like pits, etc.

At this point, I'd feel better about magnetic retention (on a hard drive) than a burnt CD/DVD, though I'd take a pressed CD/DVD over either (especially if you aren't mucking around with it and scratching it up and whatnot).

I don't think there's a silver bullet for backup and archival at this point. You need to try a lot of different ways, depending on your circumstances. For my home use, I just make sure I have data on redundant servers with mirrored drives on the servers themselves. For the most part, I'm probably good. If my house gets hit by lightning and the surge protectors fail and the harddrives essplode, I'm screwed. If my house burns down, I'm screwed.

But, if a single drive fails, I have a backup on that particular system and I know I need to make sure the data between systems is (mostly) synchronized.

Of course, things I thought were absolutely essential to be backed up for all eternity 3 or 4 years ago no longer means anything at all to me, so I'm not sure this entire process is even worth it for me anymore. I'm sure the next time I go on a power-saving kick I'll end up powering one of the file servers down, or maybe just get a few large harddrives and shove them in my main PC and power them both down.

But I digress. At length.

Re:I highly recommend using Archival Grade Media

[Chaos Incarnate]

"Burned" CDs aren't physically carved; their dye layers can change depending on how they're made and the conditions that they're subjected to.

Re:I highly recommend using Archival Grade Media

[Chabil Ha']

I have three methods of backup for the most important files (~100GB worth).

1. Burn to disk
2. Store on server
3. Archive online

If that were still not good enough, bigger, more 'important' things are going on in the world to have to worry about where my data is.

Re:The tools are called

[hairyfeet]

If you want something fast that will tell you if there is any trouble reading the disk go here, or if you want something more complex that'll go by sector go here.

Oh and in the future if the software needed can be Windows software,allow me to suggest Freeware World Team. FWT have hands down the best freeware search engine I have ever seen. You simply type into the search box what you want the software to do and they'll find you a piece of freeware that does it. I use it here at the shop all the time when myself or a customer has a job that needs filling. No spyware,no trialware,just great freeware with a truly great search engine to find it with.

Re:I highly recommend using Archival Grade Media

[Yewbert]

You're getting a couple different processes confused - writable-once and re-writable media are two very different things.

ReWritable discs ONLY (CD-RW, DVD+/-RW), use a layer of a metal alloy that undergoes a reversible phase change (crystalline/amorphous) when written (heated and cooled by different amounts at different rates). This phase change produces a very small change in reflectivity, to allow reading.

Recordable (CD-R, DVD-R, DVD+R) media have an organic dye layer that is burned irreversibly. The change in reflectivity produced when lasering that dye layer against a reflective layer is a greater change than with phase-change alloys.

Both Recordable and ReWritable CD media are also hampered by a very thin top protective layer. Microwave a coastered CD-R for a few seconds till it flashes and flakes, and see just how thin and fragile this layer is.

DVD-R/RW media is much more robust - the recordable layer is sandwiched in between two plastic layers, so I'd expect the lifespan of recorded DVDs - even based on the exact same dye - to be considerably longer than recorded CDs.

The question of whether Recordable or ReWritable media has a longer life span is one I haven't seen explored very well at all (though, and I think this was your main point), glass-mastered/stamped CD- and DVD-ROM media certainly do last longer than either.

One odd but explicable trend in Recordable media aging is that the shelf-life of a dics once written is longer than that of a blank disc. Don't get too enthused about stocking up on cheap CD-Rs or DVD-/+Rs at a sale - if you leave them sitting on a shelf too long before burning them, they will go unreliable.

shop much?

[way2trivial]

http://www.yuden.co.jp/us/product/pdf/mdvd_e.pdf

page 5, 4th col... see where it says 8.5GB? see at the top where it says DL?