Atomicity Consistency Isolation Durability

Atomicity

  • “All or nothing”
  • If any part of transaction fails , everything gets rolled back

Example: Bank transfer between 2 users at an example , if something fails every step of the transaction is rolled back

Consistency

  • Transaction must follow all rules and constraints
  • DBMS will return a consistency violation and cancel the transaction

Isolation

  • How concurrent transactions affect consistency
  • Isolation makes it seem like each transaction is the only one occurring

Different levels of Isolation (in order of allowing most consistency):

  1. Serialisable
  2. Read Committed Isolation
  3. Repeatable Isolation

Tradeoff using a higher isolation means more consistent transactions, but mean less concurrent transactions

Durability

  • Once committed, the transaction is permanent
  • Done by write-ahead logging
    • Persist changes to disk before confirming the commit
  • In distributed databases this means replicating across other nodes