Linearizability Vocabulary

1. A distributed database guarantees that 'reads always reflect the latest write' — any read issued after a successful write will never return an older value, regardless of which replica serves the request. What consistency model does this describe?

Eventual consistency Causal consistency Linearizability (strong consistency)

2. A social media platform allows users to reply to comments they have seen. The system guarantees that if user A writes a comment and user B replies to it, all other users will see A's comment before B's reply. What consistency model is this?

Linearizability Causal consistency Monotonic read consistency

3. After a write is confirmed, users may briefly read old data from other replicas. The system promises all nodes will eventually converge to the same state, but makes no timing guarantee. A product description says: 'Our system provides eventual consistency.' What does this mean?

All reads immediately reflect writes Replicas will converge to the same state given no new updates, but reads may temporarily return stale data Writes are confirmed only after all replicas are updated

4. In an architecture review, a senior engineer says: 'Linearizability is the strongest single-object consistency model, but it comes at a cost.' What is the primary cost of providing linearizability in a distributed system?

Increased storage usage on each replica Higher latency and reduced availability — achieving linearizability requires synchronous coordination between nodes before confirming a read or write More complex client-side code

5. A team evaluating databases for a financial ledger says: 'We need linearizability for balance reads — we cannot show stale balances.' For a recommendation feed feature on the same platform they say: 'Eventual consistency is fine here.' What principle guides these two different choices?

All data in a system must use the same consistency model Consistency model trade-offs — choose stronger consistency where stale data causes correctness problems, and weaker models where staleness is acceptable to gain lower latency and higher availability Financial data always requires eventual consistency to ensure scalability

Vocabulary Reference

Term	Meaning
linearizability	Every read returns the result of the most recent write; operations appear to execute instantaneously and in a single global order. The strongest single-object consistency model.
causal consistency	Operations that are causally related (one happens-before the other) are seen by all nodes in the same order; unrelated operations may appear in any order.
eventual consistency	All replicas converge to the same state given no new updates; reads may temporarily return stale data.
reads reflect the latest write	A phrase describing linearizability — no read ever returns a value older than the most recently confirmed write.
our system provides	A common phrase in architecture docs for stating a consistency guarantee: "our system provides linearizability / eventual consistency / causal consistency".
consistency model trade-offs	Different consistency models offer different latency, availability, and correctness guarantees; the appropriate choice depends on the data's correctness requirements.

Vocabulary Reference

Exercise complete!