Advanced Distributed Systems

During a network partition you choose between consistency and availability — and the answer is per-component, per-moment.

7 min 1 diagram practice

CAP's rarely-quoted second half: even without a partition, you choose between latency and consistency (Else Latency vs Consistency).

6 min practice

Linearizable behavior: a read after a write always sees the write — the default you assume with one database.

6 min practice

Replicas converge eventually — the default of any replicated system that doesn't synchronize every read.

6 min practice

The precise form of strong consistency — operations take effect at a single point in real time.

6 min practice

W + R > N: read the majority so at least one overlapping node has the latest write.

7 min 1 diagram practice

Exactly one node acts as leader at any time, others follow — via leases, fencing, and a consensus-safe tiebreaker.

7 min 1 diagram practice

The problem behind 'everyone agrees on the same order': Raft and Paxos, and why you almost always use etcd/ZK/Raft behind a library.

6 min practice

The lease + fencing story done properly — and when a lock is the wrong tool (idempotency often beats locks).

6 min practice

Two-phase commit's prepare/commit protocol — and why production prefers saga/outbox for everything except narrow windows.

6 min practice

Store the history of changes as an append-only event log — the state is derived, never stored as a mutable row.

7 min 1 diagram 1 code practice

Separate the read model from the write model — different schema, different store, different scaling.

6 min 1 diagram practice

The pattern library you choose from per flow: outbox, saga, event sourcing, CQRS, TTL caches, read-your-writes, idempotency.

6 min practice