Back to roadmap
Module 6 · Relational Data at ScaleDay 04925 min

Relational Basics and ACID

Why SQL still wins for most transactional work.

← PreviousNext →
Day 049

Relational Basics and ACID

Atomic
note
Consistent
note
Isolated
note
Durable
note
Memory hook

Relational Basics and ACID: why sql still wins for most transactional work

Mental model

shape data so reads and writes stay honest

Design lens

Higher isolation costs locks/aborts.

Recall anchors
AtomicConsistentIsolated

Why it matters

Relational databases provide ACID transactions: atomic all-or-nothing changes, consistent (invariant-preserving) writes, isolated concurrency, and durable on commit. They're the right default for any transactional workload.

Deep dive

Atomicity: a transaction either fully applies or doesn't.

Isolation levels: read-uncommitted (rare), read-committed (default in many), repeatable-read, serializable.

Most non-trivial concurrency bugs come from picking too weak an isolation level.

Demo / scenario

Transfer money between accounts.

  1. BEGIN; UPDATE A -100; UPDATE B +100; COMMIT;
  2. Atomicity ensures both or neither.
  3. Repeatable-read prevents non-repeatable reads of A.
  4. Serializable prevents anomalies even under concurrent transfers.

Tradeoffs

  • Higher isolation costs locks/aborts.
  • Most apps default to read-committed; raise where needed.
  • Optimistic concurrency control is an alternative.

Diagram

Atomic
Consistent
Isolated
Durable
ACID and isolation.

Mind map

Check yourself

Loading quiz…

Sources & further reading

← PreviousNext →