High Level Design
Consistency models (linearizable, eventual, causal, quorum), database transactions, and the four isolation levels — from read uncommitted to serializable.
Primary-replica replication, WAL and CDC, async vs sync vs semi-sync trade-offs, split-brain problem, write amplification, and zero-downtime migration strategies.
Two-Phase Commit, how Paxos achieves consensus through majority agreement, and the multi-server write coordination protocol with log-line locking.
How consistent hashing solves the massive-rehashing problem in distributed systems — hash rings, virtual nodes, load imbalance, and real-world use cases.
End-to-end design of a distributed key-value store — CAP theorem, consistent hashing, replication, conflict resolution with vector clocks, failure handling, and the detailed read/write path.
Designing a distributed unique ID generator — comparing multi-master replication, UUID, ticket servers, and Twitter Snowflake, with a deep dive into the 64-bit Snowflake layout and follow-up considerations.