Level 5 of 13 · about 3 weeks
Transactions, consistency and consensus
What the database promises, and what breaks when machines and networks fail.
Modules
Core: everyone needs it. Common: most jobs use it. Specialist: optional depth, safe to skip on a first read.
1
Transactions and ACID
What atomicity, consistency, isolation and durability promise, and what they do not
Core 6 parts · about 3 hours
2
Isolation levels
Dirty reads, read skew, lost updates, write skew and phantoms, and which level prevents which
Core 7 parts · about 4 hours
3
Serializability
Three ways to make concurrent transactions behave as if they ran one at a time
Common 6 parts · about 3 hours
4
Faults, networks and clocks
Partial failure, timeouts, unreliable clocks, process pauses and fencing tokens
Core 6 parts · about 4 hours
5
Failure detection and leader election
Pings, heartbeats, leases and the election steps of Raft
Common 6 parts · about 3 hours
6
Consistency models
Linearizable, sequential, causal and eventual, and what each one costs
Core 7 parts · about 4 hours
7
CAP and PACELC
What the CAP theorem says, what it leaves out, and the everyday cost of strong consistency
Core 6 parts · about 3 hours
8
Consensus
Getting machines to agree, the Raft log idea, and what coordination services do
Common 6 parts · about 4 hours
9
Distributed transactions
Two-phase commit, its blocking problem, sagas and the outbox pattern
Core 7 parts · about 4 hours