Independent Practice and Transfer

Complete the prerequisite check, concept sequence and worked examples first. Keep the solution page closed until you have written an attempt. These cases deepen the existing practice and lab evidence; they do not replace the original topics, project or practical gates.

Case A — Predict and justify

Two candidate architectures have the same p95 latency on successful requests, but one drops 20% of requests. Can that percentile select the better design?

Record the expected result or state transitions, the assumption that makes them valid, and one plausible but incorrect answer with a counterexample. A result without its reasoning is incomplete.

Case B — Change the assumption

Change the stimulus from steady traffic to a burst while keeping the response target. What evidence must be rerun?

Identify exactly which part of Case A survives and which must be recomputed or qualified. Keep both attempts; do not overwrite the first with the calibrated answer.

Experiment — Connect the claim to an actual observation

Use the quality-evidence evaluator, then collect actual service measurements under two tactics. Preserve scenario, stimulus, environment, response, denominator and limitations in one decision packet.

Use the existing guided labs and their setup/contract instructions. Save the exact version or source revision, input, command/procedure, predicted outcome, actual output and interpretation. Add a negative control that would expose the particular mistaken claim, then restore the correct setup and rerun. Use a disposable copy for deliberate failures.

If this task requires a runtime or reviewer you do not have, mark that observation pending. A written prediction, simulator, reference implementation or generated review cannot close an actual-system or human-evidence requirement.

Review, repair and transfer

Compare your attempt with separate practice calibration. Repair the smallest misunderstanding and answer a fresh variant unaided. At the module's existing delayed recheck, change a value, failure boundary or assumption and defend what transfers. Reuse this evidence in the module assessment; its per-gate thresholds and critical errors remain authoritative.

Count this work inside the original practice, lab and recheck allocation in semester workload and progression. Reuse a matching original exercise artifact and extend it with the changed case instead of repeating identical work. Record actual time and overruns for pilot calibration; the published estimate is not a deadline.