Slashdot Mirror
Up To 10% of CD-Rs Fail Within a Few Years
Whatever you think about the likelihood that a new kind of DVDs could last for 1,000 years, this note from reader crazyeyes should give you pause about expecting current CD-Rs to be reliably readable for decades. TechARP found a failure rate near 10% for CD-Rs recorded 7 to 9 years ago, after storage in ideal conditions. On some, one or more individual files could not be recovered; others were not reliably readable on two separate drives. "In the past, hard disk drives were small (in capacity) and costly. To make up for the lack of affordable storage, many turned to CD-Rs. As it became common to store backups and personal pictures, videos, etc. on CD-Rs, the lifespan of these discs became a concern. According to manufacturers, CD-Rs should last for decades. Some even quoted an upper limit of 120 years based on accelerated aging tests! That sure is a long time, isn't it? But will CD-Rs really last that long?"
According to manufacturers, CD-Rs should last for decades.
According to their marketing dept., rather.
I've experienced this myself lately with a bunch of disks that were now useless. It was cheaper off brand disks that failed. The irony is at the time I got them, they were the ONLY disks I could get to work on my CD player.
So far I've had no failure with CD-R's from Sony, TDK etc... Which were the disks my CD player simply would NOT play.
i have entire 1995 to 1998 CD-R spindle's and all 400 of them still function just fine. i recently had to run trough all 400 of them, and had zero read errors. i guess my discs are possessed by some magical force, or this is just bogus.
I've had CD-Rs and DVD-Rs that I burned over a decade ago still read fine. However, those disks were verified burns where I immmediately read back the data with Nero to make sure they were ok.
There was a time when I didn't do verified burns. Those disks have a ridiculously high failure rate, but I'm betting they were bad burns in the first place. With most media I get close to a 10% failure rate on verifying the burns.
I do not fail; I succeed at finding out what does not work.
I recently tried reading a bunch of audio CD-Rs burned between 2003 to 2007. I used Exact Audio Copy on a Toshiba drive. I was able to get error-free reads from about half the disks recorded in 2003; about 3/4 of the ones from 2004, and from all the ones recorded after that. On the early ones that worked, sometimes EAC took a couple of hours to do the reads, which means it was doing a lot of retries. On the later ones, the transfers were mostly just a few minutes. On the ones that reported less than 100%, sometimes EAC spent 50-60 hours trying.
For the disks that I could not get 100% reads on from the Toshiba drive, I tried them in several other computers using a variety of programs. Mostly I was not able to get results as good as EAC on the Toshiba drive. I tried two Mac Mini's using Max and an old Mac G3 using cdparanoia from the command line, and got lots of failures. Then I tried Max on my MacBook and they all read perfectly. Go figure.
I theorize that one reason the disks had errors was that they were labeled using a Sharpie. According to the NIST report on CD-R failures (nvl.nist.gov/pub/nistpubs/jres/109/5/j95sla.pdf), this is a really, really bad idea. Since I read that report, I've been adamant about using only water-based markers on CDs and DVDs.
I burn thousands of CDs and DVDs per week and here are some tips
- use pro grade media from Taiyo Yuden (Made in Japan) or Falcon (Made in UAE). Verbatim still makes some good media but you have to know what to look for (Datalife Plus) because they also buy cheap media and rebrand it.
- burn cd-r at 16 or 24x. 32x is ok for short-term use. Even the best discs will fail if you burn at maximum speed.
- burn dvd-r at 8x
- if you must burn dvd-r at 16x, test your quality regularly for signs of failure.
how to test the quality:
- Plextor made good drives bundled with Plextools testing software but they are no longer making their own drives. For a replacement to Plextools, see Opti Drive Control at cdspeed2000.com
I used-to make 2 CDs of every ISO, until I figured out how best to utilize PAR2.
PAR2 calculates parity information on a set of files, and writes out a file which can be used in the event that any of the files is damaged. This is quite similar to RAID-5, but PAR2 is more robust, and works on any files, not just equally sized hard drives.
Though it's no help on DVDs, CDs work GREAT with PAR2, because of their two different methods of recording. Mode 1, as all regular files are stored, reduces the amount of space available by about 12.5%, using that space for additional error correction data. Audio CDs, and Video CDs, where a single bit error isn't nearly as critical, are recorded in Mode 2, with substantially reduced error correction, but about 100MBs more usable space available.
PAR2 is similarly resilient to errors, so it can safely be used with Mode 2. This allows much more space for the parity information, and the opportunity to be safe against, and correct, respectively more damage to a disk.
Specifically, I recomend a 4-disk parity set. You fill 3 CDs full of data, and tell PAR2 to calculate 37% recovery data on those files. The first 33.334% allows you to RECOVER THE DATA FROM ONE COMPLETELY LOST CD, no matter which of the 3 it is. That still leaves you with a margin of 3.667%, so those two CDs you DO have, can have a few bad sectors as well, and all the data from the lost CD, as well as undamaged versions of the files on the two lightly damaged CDs can be recovered. Alternatively, if you DON'T lose an entire CD, all three (4 actually) CDs can have numerous bad sectors, in any distribution, up to a total of 37% of all the discs, and pristine data can still be recovered.
The method to do all this is quite simple. Just run the par2create command, telling it to create 37% recovery information. Then take the resulting BASENAME.Par2+??????? file, and create a CUE file, describing a CD with a single track across the whole CD, with the PAR2 file as the supposed audio data. eg.:
Now, any CD recording software that understands CUE files will happily record this to disc. On Unix systems, you can choose cdrecord, or cdrdao.
Now, like regular audio CDs and Video CDs, you can't just use or copy this data off the disc like a normal file on a CD. There are programs for converting VCDs into regular files, something like dat2mpeg, but I prefer a more generalized tool that can do the job:
mplayer vcd://2 -dumpstream -dumpfile par2.bin
You'll note that checksums of the file and the data on disk don't quite match... This is because, in mode2, data MUST be padded to the block size. PAR2 files are fine with it, and the padding is silently discarded.
Something like DD_RESCUE to copy the (normal) files off the other CDs, in the event of damage, is probably necessary as well. Then, once you've got 3CDs worth of data (eg. 700MB CDs x 3 = 2100MBytes) you can run par2recover and all with be repaired, like magic.
The only footnote being that calculating the parity information isn't fast, so this method is probably slower than just recording 2 copies of every CD. Also, if you lose more than 37% of the data across all the discs, the error-free originals can't be recovered. However, I consider it more reliable than duplicate discs, if only because the odds of an error on the same sector of two discs (or one disc lost, and the backup with a few errors), seems more likely than 37% of the discs being damaged beyond hope. And as an added bonus, you save 1/3rd on your CD-R purchases.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
Just broadcast your illegal movies and ugly photos toward a large, massive body so that the signals intersect with the earth again later after traveling along space-time geodesics. You can use Sagittarius A* (black hole at the center of the galaxy) for this, but you have to remember to be there to record your 50,000 year old backup once it arrives, because it's not like the hole is your bitch.
Back in the earlier days of CDR, a "high speed recorder" was recording at a whopping 4x or so. As drive recording speeds increased, the CDRs rated for those higher speeds had to become more responsive to the laser hitting it for a shorter period of time. How do you accomplish that? One big way was spreading the dye out in a thinner layer. That's likely to have a negative effect on longevity.