Practical notes on building reliable software by Arunkumar Ganesan

Write Ahead Log: Write Intent Before State

Why a write ahead log makes crash recovery and replication safer.

Vector Clocks: Track Which Version Saw Which

A simple explanation of vector clocks and how they detect concurrent updates.

Update Propagation: Moving a Change Across Copies

How updates move from a source of truth to caches, read models, indexes, and other replicas.

Two Phase Commit: Coordinating a Commit

A practical explanation of two phase commit, prepare, commit, and why the coordinator matters.

Serialized Transactions: One Clear Order

How serialized transactions make concurrent work behave as if it ran one transaction at a time.

Raft Consensus: Leader and Followers in Plain English

A practical explanation of Raft consensus using leader election and replicated log commitment.

Quorum: Enough Votes to Trust the Answer

A simple explanation of quorum reads and writes through a three replica example.

Partition Tolerance: Keep Rules During a Split

What partition tolerance means in practice and how to design behavior for broken communication.

Ordering: Decide What Happened First

How ordering rules prevent consumers and replicas from applying events in unsafe ways.