Storage

Storage is the platform responsibility of retaining data and making it available for later use. It covers more than the device holding the bytes: the design must identify what is retained, how it is addressed, who can read or change it, and what happens after a failure.

What must survive the job?

Suppose a sales job stops after producing a report. Its temporary working data may disappear, while the accepted sales table and the inputs needed for a permitted rebuild must remain. Memory used during a calculation is not automatically durable storage. Durability concerns surviving failures; availability concerns whether data can be accessed when needed.

Files, database tables, and objects offer different interfaces. An object store such as Amazon S3 identifies stored objects with keys and can maintain versions when configured. It is one storage mechanism, not a definition of the whole storage responsibility.

Retention is a design choice

Keeping two copies does not by itself provide a recoverable backup if the same deletion or corruption affects both. Record retention, restore procedures, ownership, and the authoritative version. Test the recovery that the system promises.

Compute performs work on stored data; serving exposes a usable result. One database may provide all three roles. Separate responsibilities in the design even when a single product implements them.

References

• Amazon S3: Storage concepts