How Long Do CDs Last? It Depends, But Definitely Not Forever

Back in the day (1999) I wrote a data processing and archiving pipeline for a research satellite, and I was surprised to discover the specificity of their data archival requirements. Off the top of my head, these were decision points: * Pretty much only used "gold" CDs, rated for longer storage. * CD write speed was restricted - higher write speeds caused more write failures * The total data payload was regulated (~600 MB I think), and one of the reasons why was because of the physical impact of human handling on the outermost areas of the disc. * Storage temperature was also regulated

But ultimately, the two main data storage strategies employed were: * Physical redundancy - six copies of everything, mailed out to six physically separate facilities every week. * No archiving "forever" - Archived data was (theoretically) to be restored to contemporary storage technology whenever there was a significant step change in storage technology. At the time, another member of the lab was working on restoring old Voyager data feeds, although I have no awareness of how common this actually was.

I have a collection of old burned CDs. All, somewhere between 8-13 years old. Recently I went through and either copied the data to an external drive or tossed 'em. I was able to get all the data off of probably more than 80%. But there were a few that I couldn't get anything off. So I'd say maybe 10 years for burned disks.

My pressed disks (store bought albums and old video games from the same vintage) seem to be 100% okay unless there's scratches on them.

The manufacturer claims these discs will last 1000 years:

http://www.mdisc.com/what-is-mdisc/

Digital data is best suited for constant replication instead of long term "cold" archival. 400,000 CDs is just 320TB of data, perfectly reasonable amount to keep on live hard drives these days. And once you are handling files instead of physical discs then you can just keep changing the underlying HW as tech marches on. With checksumming and redundancy, it seems almost impossible to have bitrot eating data.

I recently ripped all my CDs to WAV files (got tired of the space all those physical discs were taking up). 800+ CDs, some of which dated back to the dawn of the digital music era in 1983. Some of them (music from Eurythmics, Dire Straits, Pink Floyd) required several passes to get a clean copy, but I was able to copy them all.

In my case, I always kept them in their jewel boxes, which were stored vertically, much like you'd store an LP. For most of their life they were kept indoors, but they also spent about 3 years in an unheated storage unit.

The fun part was finding adapters for the few 8 cm CD singles I had, so they'd play in a modern slot-loading drive.

Since the main problem seems to be oxidation, I'm surprised nobody recommends putting them in a sealed cabinet and flushing with compressed nitrogen. Seems easy and cheap.

All you really need to keep data forever is RAID (10 is nice, 1 is the minimum; for more resiliency, you can mirror the entire array a couple of times too) with a load of hot spares, perhaps with the copies stored at 4 or more sites (2+ continents). Radiation shield the lot to reduce bit flips, and if one occurs, you have a background read process that catches them and uses a simple quorum to determine the correct value; some form of ECC could also be built into either the file format or the filesystem itself. As more storage is needed and/or the distro they are using updates a major version, build a new system in parallel, replicate first, test, then remove the old one once it's all verified as good.

An interesting article about the National Archives at Kew: http://www.theregister.co.uk/2013/12/11/feature_geeks_guide_...

Cloud storage would work best, stored at multiple data-centers for extra redundancy, stored with multiple cloud services to be extra cautious.

Gold CDs claim to last 300 years and are not too expensive: http://www.amazon.com/Kodak-Gold-Preservation-Write-Once-CD-...

Here are 50 DVDs for $14 that are suppose to last 50-100 years: http://www.amazon.com/Verbatim-DataLifePlus-Printable-Record...

The 200,000 CDs at the Library of Congress could be backed up on a RAID-1 NAS, but partnering with a cloud storage service might still be a better idea, plus it would allow people to download some of the data if the author allows it.

One of the variables are aluminum vs gold on the reflective surface. I've yet to have a gold one come back unreadable or corrupted but have a number of aluminum ones that have.

Compact disc owners have noted 'black crud' as a failure mode, where the tendrils of black oxide begin to develop on the disk surface between the layers. My original Cars CD had that happen to it and it was fascinating, it seemed to have started from a defect on the back of the disk which was not noticeable until black stuff started growing there.

I remember looking for a good DVD+R brand when I was backing up my CDs to FLAC for years and everyone had settled with Taiyo Yuden by all the various audiophile and archiving obsessed people I could find online. I haven't had a problem with the DVDs I've burned 10+ years ago now (I checked recently, I read 30 without a hitch before I was satisfied), but now I'm wondering if anything has changed since that time.

When I was in college (mid 90s) it cost about $1 per blank CD. They were rated at "10 years". However, you could pick up a gold blank for about $5 each, and those were rated at "100 years". I wonder how accurate that was. I don't have ready access to either one, sadly.

Half as long as you think they will. No exceptions.