Distributed systems layer
Distributed systems
Design services that stay correct when messages are lost, late or repeated.
- 01
Clocks lie
Why two servers disagree about the time.
- 02
Idempotency
Make a retry safe, so retrying is always an option.
- 03
Exactly once
What it really means, and what it can't.
- 04
Queues and Kafka
Ordering, partitions, and consumer offsets.
- 05
The outbox pattern
Saving data and sending an event without losing either.
- 06
Sagas
Multi-step work with a way back.
- 07
Consensus with Raft
How a group of servers agrees on one log.