Skip to main content

Book Exercise Lanes

This module's exercise system is local-chunk-driven. Use these chunks for targeted volume after you have already learned the concept from the guide. The four lanes align to the module's clusters.

How To Use This Page​

  1. Finish the relevant concept page first.
  2. Run the drill at the end of that concept page from memory.
  3. Only then open the matching exercise lane.
  4. Keep a mistake log tagged with common errors such as jumped to solution, no numbers, wrong partition key, ignored hot partition, assumed AZ does not fail, trade-off unstated, missed SPOF, over-ran phase 1.

Lane 1: Framing, Estimation, and Hard Parts (Cluster 1)​

Use this lane when your main problem is turning vague prompts into defensible numbers and a hard-parts list.

Target outcomes:

  • 5 problems framed into functional + non-functional + constraints + estimates, each in under 15 minutes
  • 5 back-of-envelope calculations done on paper without a calculator
  • 5 prompts with a 2-3 item ranked hard-parts list
  • memorized the powers-of-two row (10, 20, 30, 40) and the four core Jeff Dean latency numbers

Lane 2: High-Level Design, Storage, and Caching (Cluster 2)​

Use this lane when you can frame a problem but still reach for the same storage and caching shapes by habit.

Target outcomes:

  • 4 box-and-arrow diagrams drawn from scratch, each with 6-12 labeled boxes
  • 6 seven-line storage sketches across different workloads, with a rejected alternative named each time
  • 4 five-line cache/CDN/LB justifications per diagram
  • 1 anti-pattern hunt in one of your own diagrams, with a fix applied

Lane 3: Deep Dive -- Components, Data Model, Concurrency (Cluster 3)​

Use this lane when you can draw the diagram but freeze when a reviewer says "deep dive on this box".

Target outcomes:

  • 6 five-question decompositions on random boxes from your Lane 2 diagrams
  • 4 data-model sketches with primary key, partitioning key, and 2-3 indexes each
  • 4 three-line concurrency/consistency/transaction contracts per mutable table
  • 2 hot-partition fixes (time-bucket, read-fanout threshold, or similar) worked through end-to-end

Lane 4: Stress Test, Communication, and Doc (Clusters 4 and 5)​

Use this lane when the design is plausible on paper but you cannot defend it under scaling or failure pressure.

Target outcomes:

  • 4 full 10× and 100× walks on different designs, each producing a ranked list of structural changes
  • 4 per-box failure tables (impact, blast radius, recovery, TTR), with AZ and region rows
  • 2 ranked bottleneck lists and 2 ranked SPOF lists, with accept-with-reason sentences written
  • 1 full design doc produced (sections 1-12 from Cluster 5 concept 15)
  • 1 timed 45-minute solo interview run that hits all four phases

Self-Curated Problem Set​

Build a custom set from SDP 24: System design interview questions with these minimums:

  • 3 read-dominated systems (URL shortener, feed, content CDN)
  • 3 write-dominated systems (chat, metrics pipeline, bulk ingest)
  • 3 strongly-consistent systems (booking, inventory, payments)
  • 3 real-time systems (rate limiter, leaderboard, live scoreboard)

Rotate through them. Notice which shapes you habitually reach for and deliberately pick different ones.

Completion Checklist​

  • Completed at least two lanes in full
  • Logged at least 12 real mistakes and corrections across sessions
  • Wrote at least 10 full trade-off statements in the four-slot form ("A over B because C, accepting cost D")
  • Ran at least 3 timed 45-minute sessions, at least one scoring "hit every phase on time"
  • Produced at least 1 full design doc with all 12 sections
  • Reattempted at least 4 failed estimation problems after review