Eventual Consistency

• Eventual Consistency
  • CAP Theorem Addendum
  • Causal Consistency
    • Causality vs Linearizability

• Properties of Eventual Consistency
  • Does not require waiting for network communications
  • Causal broadcast can disseminate updates
  • Concurrent updates
    • Leads to conflict potential

CAP Theorem Addendum

• A system can be strongly Consistent (linearizability) or available, when a network is partitioned
• Can be both consistent and available, if we discount network partitions
• For instance a Node who is partitioned, must either wait indefinitely, or return stale values

Causal Consistency

• Causality is important
  • e.g.
    • Question before Answer
    • First Create than Update
    • etc.
• A system who obeys causal ordering is causally consistent

Causality vs Linearizability

• Linear
  • Total Order of Operations
  • System behaves as if a single copy of data exists
• Causality
  • Partial Order of Operations
  • We cannot say which concurrent operation happened first
• Linearizability implies causality
  • Causality does not imply linearity
• Causal Consistency is the strongest possible consistency model that…
  • can handle network delays
  • remains available in the face of network partitions

---