Day57: How to Extract Scoring Points from Case Questions and Define Quality‑Attribute Scenarios

Change the Learning Method

After completing Day01‑Day56, stop using the following habits:

Read notes from start to finish;
Assume you know a term when you see it;
Memorise a whole generic template;
Write all knowledge without checking the question;
List only "cache, cluster, message queue" as answers.

The second pass requires four actions:

Case question → circle problems, numbers, limits, objects;
Identify knowledge shelf → pull relevant tools;
Combine business context → write method, effect, cost or verification.

Pass criterion: write an answer independently first, then correct it according to scoring points.

Reference Materials

Day13 case‑analysis template: circle problem, pick pattern, attach scenario, and "technology + business object + effect".

Day24 quality‑attribute six‑element scene: stimulus source, stimulus, environment, product, response, response metric.

Day26 architecture design method: drivers, constraints, decisions.

Day56 software fault tolerance and overall review.

Official tutorial Chapter 8: system quality attributes and architecture evaluation.

202511 Architecture Designer Red‑Book (PDF) for knowledge‑gap reference.

Three‑Pass Reading Method

First Pass – Read the Questions

Quickly scan each question to determine the required output format.

Second Pass – Circle Four Types of Evidence

Problem : current pain points (e.g., slow response, single‑point failure, hard to modify).

Number : measurable target or current state (e.g., 10000 concurrent, 2 s, 5 person‑days, 30 s).

Constraint : non‑changeable boundary (e.g., must use government cloud, keep original identity platform).

Object : business object the answer must address (e.g., appointment slots, registration order, hospital interface, patient data).

Third Pass – Link Evidence to Each Question

Performance question → find concurrency, response time, DB pressure, sync link;
Modifiability question → find new feature, modification time, impacted modules, heterogeneous interfaces;
Availability question → find failure scope, recovery time, redundancy, switch requirement;
Security question → find sensitive data, access roles, encryption, audit requirement.

Skip any paragraph unrelated to the current question.

Identify Quality Attribute Using Six‑Element Scene

Transform a simple attribute name into a testable requirement by filling the six elements. Example for performance:

Stimulus Source : Patients

Stimulus : 10000 patients submit registration simultaneously

Environment : Weekday 8 am peak

Product : Registration service, slot query, registration order, database

Response : System validates, queries slots, creates order and returns result

Response Metric : 95% of requests finish within 2 s under 10 k concurrency

Only writing "the system should have high performance" is insufficient; the six elements provide concrete, measurable targets.

Scoring Sentence Template

For [problem], adopt [solution], through [business‑specific action], thereby achieve [desired effect].

Bad answer: "Use cache, message queue, cluster." Good answer includes business object, mechanism, and effect.

Common Pitfalls

Empty answer – no details.

Answering the wrong question – e.g., writing a micro‑service split when the question asks for quality‑attribute identification.

Template without business mapping – list of technologies without linking to the problem.

Inventing numbers not present in the stem.

Timed Case: Regional Medical Collaborative Registration Platform

Case description : A city builds a platform integrating 30 hospitals with heterogeneous protocols. Adding a new hospital now costs 15 person‑days and requires core code changes; target ≤5 person‑days with minimal core changes. Peak load: 10 000 concurrent patients, 95% of requests must return within 2 s, current DB CPU > 90% and some requests take 6 s. Availability: a single‑instance failure caused a 20‑minute outage; target automatic fail‑over and recovery within 30 s. Security: patient data is sensitive, role‑based access, encryption, and audit logging are required.

Question 1 – Identify Four Quality Attributes

Performance : 10 000 concurrent patients, 95% ≤ 2 s; DB CPU > 90% and 6 s latency.

Modifiability : Adding a hospital now needs core code change and 15 person‑days; target ≤5 person‑days with minimal core impact.

Availability : Single‑instance failure caused 20‑minute outage; target automatic switch and 30 s recovery.

Security : Sensitive patient data, role‑based least‑privilege access, TLS for data in‑flight, encrypted storage, detailed audit logs.

Interoperability is also relevant but not counted among the four primary attributes.

Question 2 – Two Six‑Element Scenarios

Performance Scenario :

Stimulus Source: Patients

Stimulus: 10000 patients submit registration simultaneously

Environment: Weekday 8 am peak

Product: Registration service, slot query, registration order, database

Response: System validates, queries slots, creates order and returns result

Response Metric: 95% of requests finish within 2 s under 10 k concurrency

Modifiability Scenario :

Stimulus Source: New hospital or maintenance staff

Stimulus: Integration of a new hospital interface, protocol, data format, and department code

Environment: Normal platform iteration and integration phase

Product: Hospital‑access layer, unified interface, adapters (core registration logic unchanged)

Response: Adapter added, protocol/format/encoding conversion completed

Response Metric: Integration completed within 5 person‑days without core logic changes

Question 3 – Architecture Improvement Measures

Performance :

Target problem: high DB CPU, slow response due to serial processing.
Solution: Cache read‑heavy data (doctor schedule, slot availability) in Redis; async‑process SMS and logs via a message queue; deploy multiple stateless instances behind a load balancer; optimise slow SQL and indexes.
Effect: Increase throughput and meet the 95% ≤ 2 s target.

Modifiability & Interoperability :

Target problem: core code changes required for each new hospital.
Solution: Define a unified hospital‑access contract and internal data model; build per‑hospital adapters for protocol/format/encoding conversion.
Effect: Core registration logic remains unchanged and new‑hospital integration stays within 5 person‑days.

Availability :

Target problem: single‑instance failure caused 20‑minute outage.
Solution: Deploy multiple registration instances with health checks; use load balancer to auto‑remove failed instances; conduct fail‑over drills for DB and cache.
Effect: Automatic switch and recovery within 30 s.

Security :

Target problem: sensitive patient data, role‑based access, audit requirements.
Solution: Identity authentication with least‑privilege role controls; TLS for data in‑flight; encrypt stored sensitive fields; record detailed audit logs (principal, time, object, operation).
Effect: Prevent data leakage, enforce access control, and provide traceability.

Scoring Table (Self‑Training)

Question 1 – Quality Attributes: 8 pts (1 pt per attribute name + 1 pt per evidence).

Question 2 – Performance Scenario: 6 pts (1 pt per six‑element).

Question 2 – Modifiability Scenario: 6 pts (1 pt per six‑element).

Question 3 – Four Measures: 8 pts (2 pt per measure: problem, solution, method, effect).

Answer‑Question Alignment: 2 pts (clear numbering, no off‑topic answer).

Total: 30 pts, ≥ 24 pts passes.

Hard Thresholds & Error Attribution

Missing evidence → Question 1 fails.

Missing response metric → Question 2 fails.

Only listing tech terms without business context → Question 3 fails.

Typical error sources: wrong attribute name, misplaced six‑element order, ignoring "technology + business object + effect" rule, mismatched solution, overly long un‑numbered answer.

Quick‑Check Q&A (Key Points)

Read the question first to know required output.

Circle four evidence types (problem, number, constraint, object).

Determine if a new hospital interface is a modifiability issue (requires constraint of ≤5 person‑days and minimal core change).

“System should respond quickly” is incomplete without stimulus, environment, and metric.

Scoring sentence must contain problem, solution, method, effect.

Organise measures with numbering and one main idea per point.

Five‑step answer chain: read question → three‑pass reading → map evidence to knowledge shelf → write “problem‑solution‑method‑effect” answer → verify against scoring points.

Two‑Minute Teaching Script (Process Overview)

Explain the five steps from picking up a case to correcting it: (1) read the question to determine output type and count; (2) three‑pass reading – first pass reads questions, second pass circles problem/number/constraint/object, third pass links evidence to each question; (3) map circled evidence to the appropriate knowledge shelf; (4) compose answers using the "problem‑solution‑method‑effect" template; (5) compare answer with scoring criteria and rewrite the weakest part only.