Nine short chapters covering what system-design interviews actually test. They are written for the
moment an interviewer asks “why?” or “what is that?” — mechanisms, trade-offs and the numbers
that change decisions , not product tours. The reference solutions link
into these pages instead of re-teaching.
#
Chapter
Read it for
1
The interview method the twelve stages and their clock, what each proves, the four-sentence close
2
Estimation the number kit, latency orders of magnitude, numbers that kill designs
3
Storage & partitioning picking a store from the access pattern, partition keys, hot keys, indexes
4
Replication & consistency leaders, quorums, consensus, consistency models, CAP/PACELC, clocks, conflicts
5
Caching cache patterns, invalidation, stampedes, hot keys, hit-ratio economics
6
Messaging & streams queues vs logs, ordering, delivery semantics, the outbox, backpressure
7
Correctness under failure the unknown outcome, idempotency keys, state machines, sagas, reconciliation
8
Reliability & operations SLIs/SLOs, overload control, blast radius, observability, cost
9
Real-time & connections WebSockets vs push, the connection tier, reconnect storms, cursor sync
Concept
Chapter
Decisive in
Fan-out on write vs read
5, 6
SD-08 News feed
Store-before-deliver, cursors, sequence numbers
6, 9
SD-09 Chat
Quorums, consensus, PACELC, conflict resolution
4
SD-15 Distributed KV store
Token bucket, local vs shared state, fail-open
8
SD-25 Rate limiter
Idempotency keys, sagas, double entry, reconciliation
7
SD-44 Payments & ledger
Before a first round , read chapters 1 and 2 — the method and the numbers are worth more than
any single technology.After a round , the report names your weakest dimension; read the chapter that covers it, not the
model answer to the question you just did.Before a loop , re-read the Traps at the end of each chapter — they are the claims that break
under a follow-up question.
Table of contents