Glossary
Cold storage
Cold storage is the tier for data that is rarely read but has to be kept, for compliance, reference or possible reuse, at the lowest cost per terabyte. It trades speed of access for price. Its full cost depends as much on how and how often data comes back as on the monthly storage rate.
Access frequency defines cold data
Data is cold because it is seldom read, not because it is old. A compliance copy of financial records is cold from the day it is written, while a year-old dataset retrained every week belongs in hot storage. Age is easy to measure, so policies often use it, and then find that some old data is requested constantly while some recent data is never touched. Access statistics give a more reliable classification.
Three common forms
| Form | Time to first byte | Characteristics |
|---|---|---|
| High-capacity disk, usually object storage | Milliseconds | Online at all times; higher running cost than tape |
| Tape libraries | Minutes | Lowest cost to retain; cartridges are located, mounted and read in sequence |
| Cloud archive tiers | Minutes to hours, by tier | Retrieval and egress fees; minimum storage durations |
Disk-based cold tiers keep data searchable and immediately restorable, which matters when the timing of a recall cannot be predicted. The cloud options are covered in more detail under cloud archive storage.
Where the cost appears
Keeping cold data is cheap; getting it back is where cost and delay concentrate. Cloud archive tiers charge per gigabyte retrieved and often per request, which weighs heavily on millions of small objects, and they add egress fees when data leaves the provider. Most apply a minimum storage period, so data deleted or moved early is billed as if it had stayed. Tape adds time, since each cartridge is fetched, mounted and read before data flows.
A legal request for seven years of records with a two-week deadline shows the effect. For years the storage line looked small. The recall then brings retrieval fees, bandwidth and days of staged restores into a single quarter. Modeling a full recall alongside the monthly rate is what makes cold options comparable.
Moving data into the cold tier
Data reaches cold storage through policies that act as access falls, described under storage tiering. Copies that may be needed quickly, such as recent backups or datasets likely to be analyzed again, often stay on low-cost online disk rather than an offline archive, because immediate access is worth more than the extra saving. The same logic applied to backup copies is set out under backup storage tiers.
Data that is rarely read is also rarely checked. Media degrade, formats age and newer tape drives read only a limited number of earlier cartridge generations, so long-kept collections are periodically verified and copied forward to current media. A cold tier that has never been read back is an assumption about recoverability rather than evidence of it.
Frequently asked questions
Is cold storage the same as archive?
Archive describes a purpose, long-term retention. Cold storage describes an access pattern and the media that suit it. Archives usually sit on cold storage, but cold storage also holds data that was never formally archived.
Why is cold storage cheaper?
Slower and denser media, less power per terabyte, and in cloud services a commitment to read data rarely, which the provider prices in through retrieval fees and minimum storage periods.
Can cold data be searched?
Only if an index or catalog is kept on faster storage. Without one, finding a record means recalling the data first.