1) Xác định requirements không đầy đủ

Senior không bao giờ nhảy vào vẽ kiến trúc ngay, mà hỏi lại yêu cầu.

✔ Phải biết hỏi:

• Functional requirements?
• Non-functional requirements (latency, SLA, throughput, cost)?
• Who are the users?
• Data types?
• Scale hiện tại vs scale kỳ vọng?
• Single region hay multi region?
• Strong consistency hay event consistency?

❌ Red flag: lao vào microservices hay chọn Kafka chỉ vì cool.

---

---

2) Ước lượng tải (Capacity / Load Estimation)

Senior phải biết cách:

• QPS / RPS
• Peak vs average
• Growth factor
• Storage volume / data retention

Ví dụ nếu thiết kế URL shortener:

• Reads 10x writes?
• Write 10,000 / s
• Store 1B entries / 3 years

❌ Red flag: nói kiến trúc cloud nhưng không hề tính được con số.

---

---

3) High Availability & Failover

Các câu hỏi hay:

• Nếu DB chết thì sao?
• Nếu Redis fail thì degrade thế nào?
• Active-active vs Active-passive?
• Multi AZ vs Multi Region?

Senior cần nói được:

✔ Health check, circuit breaker, auto failover, leader election, quorum (Raft/Paxos).

❌ Red flag: nói “dùng cluster là xong”.

---

---

4) Scalability bottleneck

Phải nhận ra các điểm nghẽn:

✔ Cache
✔ Queue
✔ Partitioning
✔ Replication
✔ Horizontal scaling

Câu hỏi kinh điển:

Nếu load tăng 10x bạn làm gì?

Senior cần:

• Scale stateless app horizontally
• DB – sharding / read replicas
• Cache invalidation strategy
• Rate limiting

❌ Red flag: chỉ trả lời “scale server lên” hoặc “dùng microservice”.

---

---

5) Consistency vs Availability tradeoffs (CAP theorem)

Ví dụ interviewer hỏi:

• Why not use Mongo for bank transactions?
• When to use eventual consistency?
• What is read-your-own-write?

Senior cần nói:

✔ Strong consistency cho tiền
✔ Eventual consistency cho newsfeed / cart tracking

❌ Red flag: “eventual consistency là đủ cho tất cả”.

---

---

6) Data modeling + indexing

Các hệ thống e.g. Course Scheduler, Banking Card System:

✔ Chọn primary key
✔ Partition key (hot key issue)
✔ Secondary index / inverted index

❌ Senior yếu nếu:

• Không hiểu database scan vs index lookup
• Không biết composite key
• Không hiểu write amplification

---

---

7) Cache design & cache invalidation

Câu hỏi rất hay:

How do you ensure cache consistency?

Senior phải nói:

✔ TTL + lazy eviction
✔ Write-through / write-behind / read-through
✔ Cache stampede: use mutex / probabilistic early refresh

❌ Red flag: “just put Redis in front”.

---

---

8) Messaging / Queueing / Async Processing

Các bài design realtime:

✔ Kafka vs RabbitMQ vs SQS – phải biết tradeoff:

• Ordering vs throughput
• At-least-once vs exactly-once
• Replayability
• Consumer lag management

❌ Red flag: Kafka là thuốc tiên.

---

---

9) Reliability & Observability

Senior phải trả lời:

✔ Retry pattern
✔ Circuit breaker
✔ Dead letter queue
✔ Telemetry: trace, log, metrics
✔ Health endpoint + graceful shutdown

❌ Red flag: nói microservice nhưng không nhắc monitoring.

---

---

10) API contracts / versioning / idempotency

Interviewers sẽ hỏi:

• How do you design idempotent APIs?
• How do you handle partial failure?
• How do you structure versioning?

Senior cần:

✔ Request IDs
✔ Operation tokens
✔ Rollback / Saga pattern
✔ Idempotency key for payment

---

---

11) Security / Auth / Multi-tenant ( rất hay hỏi )

• JWT vs session?
• OAuth2 flows?
• RBAC vs ABAC?
• Multi-tenant isolation: schema per tenant / shared schema + tenant_key

❌ Red flag: nghĩ security = JWT là xong.

---

---

12) Storage + Database Selection

Senior phải justify:

• SQL cho consistency
• NoSQL cho scalable reads
• Time-series DB vs OLAP vs OLTP

Câu hỏi điển hình:

thiết kế hệ thống payment → bạn dùng DB gì? vì sao?

---

---

13) Transaction + Saga vs 2PC

Senior phải biết:

• Why 2PC is dangerous in distributed system
• Saga patterns: choreographed vs orchestrated
• Compensation transaction & state machine

---

---

14) Evolution & Maintainability

Interviewers test maturity:

✔ Feature toggle
✔ Backward compatibility
✔ Blue-green / canary deployment
✔ Migration strategy (e.g., schema change)

---

---

15) Cost Awareness

Câu hỏi rất “Senior”:

• If budget limited, how do you re-architect?
• Avoid over-engineering?
• Reserved instances / storage lifecycle

Senior phải biết cost trade-off.

---

---

16) Real Scenarios thường bị hỏi

⭐ A. Design an URL Shortener

Focus:

• DB sharding
• Counter / hash collision
• Cache
• Analytics pipeline

---

⭐ B. Design a Messaging App (WhatsApp / Messenger)

Focus:

• Delivery semantics
• Message persistence
• Fan-out
• Presence status

---

⭐ C. Design Payment Gateway

Focus:

• Idempotency
• Transaction safety
• Anti-fraud async process
• Dead letter

---

⭐ D. E-commerce Checkout / Flash Sale / Ticket Booking

Focus:

• Prevent oversell
• Quota management / locking
• Queue + rate limiting

---

⭐ E. Newsfeed / Social Network Feed

Focus:

• Push vs pull
• Fan-out vs fan-in
• Cache + ranking store