In 2009, the Library of Congress commissioned a research report into the degradation ofCD-ROMs in storage as a way of assessing the integrity of the media in its collection: the news isn't pretty.
The standards for certifying a CD-ROM did not "place any requirement on the chemical or physical stability of the disc," so depending on the manufacturer and process, the discs you've put away on shelves may have wildly different material properties.
The study involved taking a trove of discarded/duplicate CD-ROMs from the LoC's collection and subjecting them to "accelerated aging" processes to see how many errors emerged as the media aged. Keeping discs dry and cool helped reduce error rates, but even so there's a lot of bitrot there.
One thing that's happened since this study is an acceleration in the plunging costs of online storage -- HDDs and SSDs -- and cloud services, which are all "live" media, regulated by microcontrollers that continuously poll their storage media for degradation, marking off sectors as bad when they turn and copying their data to still-good sectors before it becomes unreadable.
This is a major difference between today's state of affairs and the long, awkward adolescence of mass storage, when keeping all your data online was prohibitively expensive, which meant that some fraction of your archives would end up on offline/nearline media, from tapes to CDs to Zip and Jazz and floppy discs. All media is subject to entropy, but offline/nearline media is not easily hedged against the Second Law of Thermodynamics with measures like continuous scheduled offsite backups and continuous defect-scanning.
The results of this study show that individual CD-ROM life expectancies in a large collection such as that held by the Library of Congress can be expected to cover a wide range. In addition, the BLER degradation rate of individual discs will be dependant on the environmental conditions to which the disc is exposed. Selecting optimal conditions for temperature and relative humidity in facilities where compact discs are stored can be expected to have a significant impact on service life.
Other factors not covered in this study, such as handling, labeling, and exposure to certain materials or chemicals, also affect service life and must be considered as part of a comprehensive approach to preserving digital information on compact disc media.
The test population selected for this experiment was extremely diverse; representing discs constructed using different materials, from different manufacturers and record labels. Although the selected discs covered a relatively limited period of manufacture the wide distribution of life expectancies demonstrates the effect of these varied construction parameters on disc life. 10% of the discs failed at an estimated life of less than 25 years, including 6 discs (5%) that failed too early to obtain meaningful data or a meaningful lifetime estimate. 23 discs (16%) had insufficient increase in errors during the test, and thus, had infinite lifetimes, by the standards of the ISO test method. These results illustrate why it is so difficult to make broad generalizations about the lifetime of optical media.
The Library of Congress plans to conduct analyses of the material composition of selected discs from both this study and the on-going Natural Aging Study to look for trends in failure modes as they relate to the chemistry of the disc. An understanding of these failure modes can help in identifying discs that are prone to early failure so that the data can be transferred to more stable media before they reach end-of-life.
COMPACT DISC SERVICE LIFE: AN INVESTIGATION OF THE ESTIMATED SERVICE LIFE OF PRERECORDED COMPACT DISCS (CD-ROM) [Chandru J. Shahani, Michele H. Youket and Norman Weberg/Library of Congress]
(via Beyond the Beyond)
(Image: CD-ROM, Black and White, CC-BY-SA)