Synchronous and Asynchronous

Core Idea

Synchronous = chờ kết quả (blocking)

So sánh tổng quan

Execution Model

Synchronous

Client → Service A → Service B → Response
                      ⏳ wait

• Caller bị block
• Phải đợi downstream xong

Asynchronous

Client → Service A → Queue → ACK
                        ↓
                   Service B xử lý sau

• Trả response sớm
• Xử lý background

1/ Synchronous

Characteristics

• Request–response trực tiếp
• Thường dùng HTTP / gRPC
• Tight coupling giữa services

Ưu điểm

• Dễ hiểu, dễ debug
• Error visible ngay
• Phù hợp transactional logic

Nhược điểm

1) Latency stacking

A → B (100ms) → C (200ms)
Total = 300ms+

2) Cascade failure

• B down → A fail → user fail

3) Limited scalability

• Thread bị block → resource waste

Use cases

• Login / Auth
• Payment
• CRUD APIs
• Real-time validation

2/ Asynchronous

Characteristics

• Event-driven
• Message queue / stream
• Loose coupling

Ưu điểm

1) High throughput

• Không block → xử lý song song

2) Fault isolation

• Service B chết → queue giữ message

3) Better UX

• Response ngay (ACK)

Nhược điểm

1) Debug khó

• Không có stack trace end-to-end

2) Eventual consistency

• Data không update ngay

3) Duplicate processing

• Cần idempotency

Communication Patterns

Sync

• REST
• gRPC

• Message Queue (RabbitMQ)
• Event Stream (Kafka)
• Pub/Sub

Consistency Model

Real System = Hybrid

Không system nào chỉ dùng 1 kiểu

Example: E-commerce

Sync

• Create order
• Payment

• Send email
• Update analytics
• Notify shipping

Pitfalls (Interview gold)

Async pitfalls

• ❌ Duplicate events → need idempotency
• ❌ Out-of-order events
• ❌ Message loss (if misconfigured)

Sync pitfalls

• ❌ Timeout
• ❌ Retry storm
• ❌ Circuit breaker needed

Design Patterns

Sync patterns

• Circuit breaker
• Timeout
• Retry with backoff

Async patterns

• Queue / PubSub
• Dead Letter Queue (DLQ)
• Event sourcing
• Saga pattern

Decision Framework

Use Sync when:

• Cần kết quả ngay
• Cần strong consistency
• User-facing critical path

Use Async when:

• Task dài / không critical
• Scale lớn
• Tolerate delay

One-liner (Interview)

“Use synchronous for critical, real-time operations and asynchronous for scalable, decoupled background processing.”

Mental Model

Sync = correctness first

Advanced Insight

Async không phải “faster”

• Thực tế:

Sync không phải “bad”

• Phần quan trọng nhất của system luôn sync