System Design Interview - An Insider's Guide by Alex Xu & Sahn Lam

Bought Vol 1 in 2020, Vol 2 in 2022. Used for every system design interview (Google, Meta, Amazon, Stripe, Databricks). The framework (Requirements → API → Data Model → High-Level → Deep Dive → Bottlenecks) is now muscle memory.

Case studies are production-realistic: TinyURL (consistent hashing), Rate Limiter (token bucket + Redis), Chat (WebSocket + pub/sub), Search (inverted index + sharding), YouTube (CDN + adaptive bitrate). Each covers trade-offs: consistency vs latency, SQL vs NoSQL, push vs pull.

Vol 2 adds: Distributed transactions (Saga), Idempotency, Leader election, Observability. The "Design Instagram" case study alone is worth the price — covers feed generation (fan-out vs pull), media storage, ranking.

Not a video course. PDF/EPUB + diagrams. Read, internalize, whiteboard. Diagrams are exportable (PNG/SVG) for interview prep notes.

Did both. Grokking is broader (50+ designs), shallower. ByteByteGo is fewer (20ish), deeper. Alex Xu explains *why* not just *what* — e.g., why Cassandra for time-series, why not Postgres partitioning.

The "Back of Envelope" estimation section is crucial. Interviewers expect you to calculate: QPS, storage, bandwidth, memory. ByteByteGo drills this: 100M DAU → 100M writes/day → ~1200 QPS average, 10k peak → storage calc → shard count.

Updates are free lifetime. Vol 1 v1 → v2 added 8 new case studies, revised diagrams. Vol 2 still getting updates (added "Design Kafka" in 2024).

Community Discord (15k+ members) — Alex and Sahn answer questions. Mock interview partners findable.

Assumes you know: HTTP, DNS, TCP/IP, databases (ACID, indexes), caching basics, load balancing, CAP theorem. If these are fuzzy, do "Designing Data-Intensive Applications" (DDIA) Ch 1-3 first, or Grokking's intro chapters.

I jumped in cold, spent 2 weeks on Wikipedia/YouTube filling gaps. Then ByteByteGo clicked.

Format: text + diagrams. No video. Some prefer video (try "System Design Interview" by Frank Kane on Udemy — follows similar framework). But text lets you ctrl+F, annotate, re-read at your pace.

Price: $39 Vol 1, $39 Vol 2, or $59 bundle (often $49 sale). One-time, lifetime, DRM-free PDF/EPUB. Best value in interview prep.

Meta system design is brutal — 45 min, zero hand-holding, follow-ups go deep. Used ByteByteGo's "Design Facebook Feed" and "Design Messenger" case studies. The feed ranking pipeline (candidate generation → scoring → re-ranking) matched exactly what my interviewer asked.

Key insight from course: always clarify *scale* first. "Design Instagram" — is it 1M users or 1B? 10 photos/day or 100? Read/write ratio? These change everything.

The "Common Mistakes" callout boxes saved me: "Don't jump to microservices," "Don't use Kubernetes as default," "Don't ignore clock skew." Interviewers probe these.

Content: 5/5. Physical book (Amazon paperback): thin paper, diagrams lose detail in grayscale, binding cracked after 2 reads. Get the digital version (color diagrams, searchable, updates).

Also: the website (bytebytego.com) has free newsletter — weekly system design deep-dives (Design WhatsApp, Design Redis, Design GraphQL Gateway). Subscription worth it alone.

Errata: minor typos in early print runs. Digital auto-updates fix this.