Hot, Warm, and Cold Storage Tiers
Storage tiers match data to media by how often and how fast it must be accessed, keeping frequent data fast and rare data durable.
Not all data is accessed equally
Freshly produced data is often read constantly; old data is read rarely but must be kept. Storing everything on the fastest media is wasteful, and storing everything on the slowest makes recent work painful. Tiered storage matches each piece of data to media appropriate to how it is used, and moves data between tiers as its access pattern changes.
The tiers
- Hot: fast media for frequently accessed, latency-sensitive data; recent shots, live dashboards.
- Warm: slower, larger media for occasionally accessed data; the last months of results.
- Cold: high-capacity, high-latency media for rarely accessed data that must be retained; the long-term archive.
- Archive: deepest, cheapest tier, retrieval may take time; permanent preservation.
Lifecycle policies
Movement between tiers is governed by lifecycle policies: rules that migrate data as it ages or as access falls off, for example moving a dataset from hot to cold after ninety days without a read. Automating this keeps the fast tier lean without a human deciding for every file.
Latency versus durability
Cold tiers trade access speed for capacity and durability. Retrieval from the deepest archive may take minutes to hours, which is fine for data that is rarely needed but must never be lost. The key is to ensure that data whose access is latency-sensitive is not stranded in a slow tier.
In a fusion program
Live diagnostic and telemetry data sits in a hot tier during and shortly after a pulse; the calibrated per-run record ages into warm and then cold storage; the deposited, checksummed published record lives in a durable archive tier engineered for permanence. This mirrors the real-time versus archival split and keeps the long-lived published record safe without slowing daily work.