Understand the decision.
System design becomes clearer when a choice has an observable consequence. TraceToScale connects architecture decisions to a recorded experiment and a reasoned explanation.
Predict, inspect, explain, transfer.
- Read a brief with explicit workload and model assumptions.
- Record a prediction. Choosing “unsure” is valid.
- Change a supported parameter and run the declared model.
- Pause, step or scrub the recorded trace and compare the evidence.
- Explain the cause, your chosen change and a downside.
- Respond to an unfamiliar requirement that challenges the previous answer.
A successful constraint check and a clear explanation are separate pieces of evidence. A completed lab is practice, rather than a guarantee of interview performance.
Twelve connected topics.
Capacity and placement
Explore request capacity, caches, rate limits, link-service design, consistent hashing, replication and identifier assumptions.
Delivery and recovery
Inspect background jobs, notification retries, feed fanout, reconnecting chat, ambiguous payment outcomes and a dependency incident.
Experiments share a versioned event model. Their metrics and visual state come from the recorded run; changing playback speed does not change what happened.
A model you can question.
Workload, service capacity, queues, retries, faults and cache behavior are declared teaching inputs. Synthetic operation credits help compare work; they are not a cloud bill. A configured capacity is not a measured database benchmark.
Each lesson states its boundaries and includes a counterexample. An unsupported architecture is identified as unsupported by the selected model. The studio compiles its supported read-path grammar into an actual experiment, and shows the parameter bindings.
Selected spatial views show regions and zones using the same recorded state as the flat diagram. A component table, stepping controls and text remain available when motion or WebGL is unsuitable.
Practice privately at your pace.
Browser notebooks and studio drafts stay on the device until an explicit account operation acknowledges synchronization. Timed interview rooms offer authored follow-ups, written notes and self-review. Their deadline comes from the server.
Optional spoken explanations are private attachments with explicit recording and upload consent. Optional AI feedback depends on service configuration and usage limits; authored evidence prompts remain useful when that provider is unavailable.
Original teaching assets and clear references.
The curriculum uses the system design notes repository as a topic reference. The project authors its own scenario wording, synthetic inputs, exercises and diagrams. Lesson provenance links primary mechanism references and records version-specific review.
Published technical review checks the declared teaching model. It does not establish production sizing accuracy, validate a learner’s employment prospects or substitute for testing a real service.