Skip to main content

Book Exercise Lanes

This module's exercise system is book-driven. Use these local chunks for targeted volume after you have already learned the concept from the guide.

How To Use This Page​

  1. Finish the relevant concept page first.
  2. Solve at least one problem of your own from memory.
  3. Only then open the matching exercise lane.
  4. Keep a mistake log with tags such as averaged percentiles, missing timeout, no idempotency key, SLO without burn-rate alert, cache stampede, sticky session assumed, mitigation confused with resolution, blameful postmortem.

Lane 1: Performance Reasoning​

Use this lane when percentile reasoning, USE/RED dashboards, or scalability-law math are still effortful.

External calibration:

Target outcomes:

  • 3 percentile drills on sample distributions (compute p50/p95/p99, explain to a peer why averages do not tell you p99).
  • 1 USE dashboard design and 1 RED dashboard design for two different real services you have access to.
  • 1 Amdahl's-Law + USL calculation worksheet for a proposed scale-out, with peak N identified.
  • 1 written refutation (in your own words) of a teammate's "our average is fine" claim, using real production data.

Lane 2: Scaling Strategies​

Use this lane when you can quote definitions but struggle to defend a specific scaling decision under pushback.

Target outcomes:

  • 1 end-to-end stateful-component audit of a real application (every hop labeled stateless/sticky/external/store).
  • 1 three-layer cache design (CDN + reverse proxy + app) for a specific product, with update pattern and invalidation plan for each layer.
  • 1 thundering-herd avoidance plan for a hot key in a real cache.
  • 1 written SWOT of sticky sessions vs external session store for a specific team's use case.

Lane 3: Reliability Engineering​

Use this lane when SLOs feel abstract, error budgets feel theoretical, or "blast radius" is a slogan.

External ladders:

Target outcomes:

  • 3 SLI/SLO docs for 3 different real APIs, each with explicit burn-rate alerts.
  • 1 classification exercise: given 10 real postmortems, label each as cascading, correlated, or gray failure.
  • 1 chaos-experiment proposal with a hypothesis, blast-radius controls, and abort conditions.
  • 1 error-budget policy written for a real team: when does the team stop shipping features?

Lane 4: Capacity Planning and Load​

Use this lane when Little's Law still requires a calculator or back-pressure feels handwavy.

External ladders:

Target outcomes:

  • 3 Little's-Law calculations on real or simulated queues, with L = λ × W compared to measurement.
  • 1 rate-limiter design (token bucket or leaky bucket) with request path, storage, and degraded-mode plan.
  • 1 capacity-plan worksheet: forecast 6 months of growth for a service, with instance counts at current, forecast-linear, and worst-case geometric growth.
  • 1 back-pressure design for a worker pool: at what utilization do you shed, what do you return, how do you log?

Lane 5: Incident and Observability​

Use this lane when you know the phases but your postmortems still come out blameful or your dashboards lie to you.

External ladders:

Target outcomes:

  • 1 instrumentation audit: for one real service, list every metric, every structured log field, and every trace span. Identify gaps.
  • 2 tabletop incident runs (from Kata 4): play IC in writing, name mitigation separately from resolution.
  • 2 blameless postmortem rewrites of public incidents: take a published blamed narrative, write the systemic version.
  • 1 action-item retrospective: pick a real team's last 10 postmortem action items; audit how many shipped.

Self-Curated Problem Set​

Build a custom set with these minimums:

  • 3 percentile drills on real data.
  • 3 SLI/SLO designs for real APIs.
  • 3 Little's-Law calculations on real queues.
  • 3 cache designs for different read/write mixes (read-heavy, write-heavy, mixed).
  • 2 rate-limiter designs for different constraints.
  • 2 chaos experiments with full hypothesis + blast radius + abort.
  • 2 tabletop incidents with written timelines.
  • 2 blameless postmortem rewrites.

Completion Checklist​

  • Completed at least one lane in full.
  • Logged at least 12 real mistakes and corrections.
  • Can compute any of L, λ, W from the other two without a calculator for small numbers.
  • Can draft an SLO doc for any real API in under 30 minutes.
  • Can name the dominant failure-mode category of an incident within a minute of hearing it.
  • Wrote at least one blameless postmortem for a real or simulated incident.
  • Designed at least one rate limiter end to end.