Hot, Warm and Cold Data
Classifying data by how frequently and how urgently it is accessed, so each tier can be stored on media priced for that access pattern.
Almost all data follows the same curve: heavily read for days, occasionally for weeks, essentially never afterwards — while retention requirements keep it for years.
Hot — accessed constantly, needs millisecond latency. Primary database, cache, standard object storage. Warm — occasional access, seconds acceptable. Infrequent-access storage tiers, columnar files, a replica rather than the primary. Cold — rare access, minutes to hours acceptable. Archive tiers, at roughly a tenth to a hundredth of hot cost.
The design work is not choosing tiers, which is easy, but making the transition automatic. A lifecycle policy on object storage, date-based partitioning in databases so cold partitions can be detached, and a query layer that can still reach the archived data when someone genuinely needs it.
Three costs to check before writing the policy: retrieval from archive costs money and time; minimum storage durations make early deletion chargeable; and per-object transition requests cost money, so tiering fifty million tiny objects can cost more than leaving them.
And the important interaction: tiering is not deletion. Data in an archive tier is still held, still discoverable, and still your liability under a retention or erasure obligation.