Notesstream🌱 seedling

System of Record

Also: Source of Truth · Authoritative Data · Primary Data · Single Source of Truth

The store that holds the authoritative version of some data: new facts are written there first, and if any other copy disagrees, the system of record wins.

Why it exists

Real applications keep the same facts in many stores (database, index, cache, warehouse). Without one agreed source, a disagreement has no answer and repairs have nothing to rebuild from.

How it works

  • It's a role, not a technology: Postgres, an event log or a Kafka topic can each be the system of record.
  • Each fact is represented once, usually normalized, as in Normalization.
  • Every other copy is Derived Data, computed from it and rebuildable from it.
  • Changes leave it in a defined order: its Replication Log, a Change Data Capture stream, or the event log itself in Event Sourcing.
  • Different entities can have different systems of record (orders in Postgres, clicks in a Kafka topic). What matters is that each entity has exactly one.

Trade-offs

Gains Costs
one place to resolve disagreements the record must be durable and highly available
derived stores become disposable and rebuildable every derived store lags it a little
clear direction of data flow writers must not update derived stores directly

In practice

  • CDC pipelines make an OLTP database the system of record and treat every downstream index and cache as a follower.
  • Event-sourced systems make the event log the system of record, and the tables become views of it.

Pitfalls

  • Two "systems of record" for the same entity, with the application writing both (dual writes): they will diverge.
  • Letting a team "fix" data directly in a derived store; the next rebuild silently undoes it.

Recall

These cards are also in your review deck, scheduled with spaced repetition.