Master production-grade backend primitives through interactive cluster topologies, live chaos drills, and zero-dependency Go contracts.
Lock down core functional APIs, consistency requirements (CP vs AP), and concrete numbers: peak QPS, P99 latency ceiling, RAM working set, daily storage growth, and replication factor.
Define exact REST/gRPC contracts, idempotency headers, sharding/partition keys, and on-disk structures (WAL frames, SQL tables, sparse indexes, or Raft metadata).
Walk cleanly through the read path and write path end-to-end: Anycast Edge → Stateless Gateways → Sharded Cache / Commit Log / ACID Store → Async Replication.
Proactively break your own architecture: state cluster partitions, hot-key saturation, thundering herds, power loss WAL replay, split-brain fencing tokens, and NTP backward clock jumps.
Contrast rejected alternatives explicitly (LSM vs B+Tree, Pull vs Push, 2PC vs Saga, Anycast vs GeoDNS) and close with alertable P99 SLO metrics.