Glossary
Hot storage
Hot storage is the storage tier for data that applications read and write constantly, built on the fastest and most expensive media in a system, usually NVMe flash. It holds the working set: a small share of everything an organization keeps, but the part that users and applications are waiting on at any moment.
What makes data hot
Frequency of access, not age, places data in the hot tier. Live transactional databases, virtual machine disks, active analytics datasets and files under edit are typical. Requests tend to be small, random and concurrent, so the tier sits on flash with latency measured in microseconds.
Sized by demand, not by terabytes
The limit a hot tier reaches first is usually request rate or data rate, not capacity. A tier can be half empty and still saturated, with requests queueing while free space remains, and adding terabytes changes nothing. Sizing starts from the peak IOPS, throughput and latency targets of the applications it serves, with capacity following. Because the tier is costly, room for growth is cheaper in the tiers below.
How idle data accumulates
Data stays where it was written unless something moves it. Over months, finished projects, files nobody reopens and old snapshots collect on the fastest media, and the tier looks full while most of it sits quiet. Expanding it then means buying more of the most expensive capacity in the system. Demotion based on access, handled by storage tiering policies, moves data down as reads stop, toward cold storage at the far end.
Not all hot data writes heavily
Read-dominated hot sets, such as training data or content served to many users, can sit on dense QLC flash at a lower cost per terabyte. Write-heavy roles stay on higher-endurance drives.
Between hot and cold sits a warm tier for data read occasionally, often on hybrid or dense flash. A hot tier can also be partly virtual: a flash cache in front of slower media serves repeated reads at flash speed without holding a full copy of the data on the fastest drives.