system-architecture-advisor（系统架构顾问）

System Architecture Advisor

Overview

Expert guidance for designing robust, scalable, and maintainable software architectures. Assists with architecture patterns, API design, tech stack selection, visual system diagrams, and best practices for modern application development.

When to use this skill:

• Designing new system architecture
• Evaluating or refactoring existing architecture
• Choosing between architecture patterns
• Planning API design (REST, GraphQL, RPC)
• Creating system diagrams and documentation
• Stack selection and technology evaluation
• Scalability and performance planning
• Security architecture design

Core Architecture Patterns

Pattern Selection Framework

Assess these factors first:

1. Scale Requirements

- Current users: <1K, 1K-100K, 100K-1M, >1M - Growth projection: 6 months, 1 year, 3+ years - Geographic distribution: Single region, multi-region, global

1. Team Structure

- Team size: 1-3, 4-10, 11-50, 50+ - Team distribution: Co-located, distributed, hybrid - Technical expertise: Junior, mixed, senior

1. Business Constraints

- Time to market: Critical, normal, flexible - Budget: Startup, growth, enterprise - Compliance: Standard, regulated industry, highly regulated

Architecture Pattern Decision Tree

Start here and follow the path:

Is this a new project or refactoring existing?
│
├─ NEW PROJECT
│  │
│  ├─ Team size 1-3, MVP needed fast?
│  │  └─→ **MONOLITH** (Layered or Modular)
│  │
│  ├─ Multiple teams, different domains?
│  │  └─→ **MICROSERVICES**
│  │
│  ├─ Event-heavy, real-time processing?
│  │  └─→ **EVENT-DRIVEN ARCHITECTURE**
│  │
│  └─ Variable workload, cost-sensitive?
│     └─→ **SERVERLESS**
│
└─ REFACTORING EXISTING
   │
   ├─ Monolith too complex, teams blocked?
   │  └─→ Gradually extract to **MICROSERVICES**
   │
   ├─ Tight coupling causing issues?
   │  └─→ Introduce **EVENT-DRIVEN** patterns
   │
   └─ Cost or scaling issues?
      └─→ Migrate components to **SERVERLESS**

Pattern Details with Pros/Cons

Monolithic Architecture

When to use:

• Startups needing fast iteration
• Small teams (1-10 developers)
• Straightforward business logic
• Predictable traffic patterns

Structure:

Layered Monolith:
├─ Presentation Layer (UI/API)
├─ Business Logic Layer
├─ Data Access Layer
└─ Database

Modular Monolith:
├─ User Module (domain-driven)
├─ Product Module
├─ Order Module
├─ Payment Module
└─ Shared Kernel

Pros:

• Simple deployment and debugging
• Easier local development
• ACID transactions work naturally
• Lower operational complexity

Cons:

• Scales as single unit (all or nothing)
• Technology lock-in
• Longer build/test times as app grows
• Team conflicts in large codebases

Example Stack:

• Next.js + Supabase: Modern, production-ready monolith
• Django + PostgreSQL: Batteries-included framework
• Rails + PostgreSQL: Convention over configuration

Microservices Architecture

When to use:

• Multiple teams working independently
• Different domains with distinct scaling needs
• Need for technology diversity
• Mature DevOps capabilities

Structure:

API Gateway
├─ User Service (Node.js)
├─ Product Service (Go)
├─ Order Service (Python)
├─ Payment Service (Java)
└─ Notification Service (Serverless)

Each service:
├─ Own database
├─ Independent deployment
└─ RESTful/gRPC APIs

Pros:

• Independent scaling per service
• Technology flexibility
• Team autonomy
• Fault isolation

Cons:

• Distributed system complexity
• Network latency between services
• Eventual consistency challenges
• Higher operational overhead

Key Patterns:

• API Gateway: Single entry point, request routing
• Service Discovery: Dynamic service location
• Circuit Breaker: Prevent cascading failures
• Saga Pattern: Distributed transactions

Event-Driven Architecture

When to use:

• Real-time data processing
• Decoupled components needed
• Asynchronous workflows
• IoT or streaming data

Structure:

Event Producers → Event Bus → Event Consumers
                    │
                (Kafka/RabbitMQ)
                    │
        ┌───────────┼───────────┐
        ↓           ↓           ↓
    Service A   Service B   Service C

Pros:

• Loose coupling between services
• Excellent scalability
• Asynchronous processing
• Real-time responsiveness

Cons:

• Eventual consistency
• Debugging challenges (event flows)
• Event schema management
• Requires message broker infrastructure

Technologies:

• Apache Kafka: High-throughput, distributed log
• RabbitMQ: Feature-rich message broker
• AWS EventBridge: Managed event bus
• Google Pub/Sub: Cloud-native messaging

Serverless Architecture

When to use:

• Variable/unpredictable traffic
• Event-triggered workloads
• Cost optimization priority
• Minimal ops overhead desired

Structure:

API Gateway → Lambda Functions → Managed Services
              │
              ├─ Auth Function → Cognito
              ├─ CRUD Function → DynamoDB
              └─ Processing Function → S3

Pros:

• Pay per execution
• Auto-scaling to zero
• No server management
• Fast deployment cycles

Cons:

• Cold start latency
• Vendor lock-in risk
• Debugging complexity
• Stateless constraints

Providers:

• AWS Lambda + API Gateway + DynamoDB
• Cloudflare Workers + D1 + R2
• Vercel Functions + PostgreSQL
• Supabase Edge Functions

API Design Guidance

REST vs GraphQL vs RPC Decision Matrix

REST API Best Practices

Resource naming:

GET    /users              # List users
POST   /users              # Create user
GET    /users/{id}         # Get user
PUT    /users/{id}         # Update user (full)
PATCH  /users/{id}         # Update user (partial)
DELETE /users/{id}         # Delete user

# Nested resources
GET    /users/{id}/orders  # User's orders
POST   /users/{id}/orders  # Create order for user

Response structure:

{
  "data": { /* response payload */ },
  "meta": {
    "page": 1,
    "per_page": 20,
    "total": 150
  },
  "links": {
    "self": "/users?page=1",
    "next": "/users?page=2",
    "prev": null
  }
}

Status codes:

• 200 OK - Successful GET, PUT, PATCH
• 201 Created - Successful POST
• 204 No Content - Successful DELETE
• 400 Bad Request - Validation error
• 401 Unauthorized - Missing auth
• 403 Forbidden - Insufficient permissions
• 404 Not Found - Resource doesn't exist
• 500 Internal Server Error - Server failure

GraphQL API Patterns

Schema design:

type User {
  id: ID!
  name: String!
  email: String!
  orders: [Order!]!
}

type Order {
  id: ID!
  total: Float!
  items: [OrderItem!]!
  user: User!
}

type Query {
  user(id: ID!): User
  users(limit: Int, offset: Int): [User!]!
}

type Mutation {
  createUser(input: CreateUserInput!): User!
  updateUser(id: ID!, input: UpdateUserInput!): User!
}

type Subscription {
  orderCreated: Order!
}

Pros:

• Single endpoint
• Client specifies needed fields
• Strong typing
• Real-time with subscriptions

Cons:

• Caching complexity
• N+1 query problem (use DataLoader)
• Learning curve for teams

Tech Stack Recommendations

Full-Stack JavaScript (Modern SaaS)

Stack: Next.js + Supabase + Vercel

Frontend: Next.js 15 (App Router)
Backend: Next.js API Routes + Supabase Edge Functions
Database: PostgreSQL (Supabase)
Auth: Supabase Auth
Storage: Supabase Storage
Hosting: Vercel (frontend), Supabase (backend)

Pros:

• Single language (TypeScript)
• Fast development
• Excellent DX
• Built-in auth, storage, real-time

Cons:

• Vendor lock-in (mitigated: open-source)
• Limited to JavaScript ecosystem

Perfect for: SaaS products, web apps, MVPs

Python Backend + React Frontend

Stack: FastAPI + PostgreSQL + React + AWS

Frontend: React + Vite + TanStack Query
Backend: FastAPI (Python)
Database: PostgreSQL
Cache: Redis
Deployment: AWS ECS/Lambda + S3 + CloudFront

Pros:

• Python for data-heavy backend
• Rich ML/data science ecosystem
• React for modern UI
• AWS flexibility

Cons:

• More moving parts
• Two languages to maintain

Perfect for: Data-intensive apps, ML integration, complex business logic

Enterprise.NET

Stack: ASP.NET Core + SQL Server + Azure

Backend: ASP.NET Core Web API
Database: SQL Server / CosmosDB
Frontend: Blazor / React
Deployment: Azure App Service + Azure SQL

Pros:

• Enterprise-grade tooling
• Excellent IDE support (Visual Studio)
• Strong type safety
• Azure integration

Cons:

• Microsoft ecosystem lock-in
• Higher learning curve

Perfect for: Enterprise applications, financial systems, compliance-heavy domains

Visual Diagrams with Mermaid

System Context Diagram

graph TB
    User[User]
    Admin[Administrator]

    subgraph "System"
        API[API Gateway]
        Auth[Auth Service]
        App[Application Service]
        DB[(Database)]
    end

    Ext1[Payment Gateway]
    Ext2[Email Service]

    User -->|HTTPS| API
    Admin -->|HTTPS| API
    API --> Auth
    API --> App
    App --> DB
    App -->|REST| Ext1
    App -->|SMTP| Ext2

Microservices Architecture

graph LR
    Client[Client Apps]
    Gateway[API Gateway]

    subgraph "Services"
        US[User Service]
        PS[Product Service]
        OS[Order Service]
        NS[Notification Service]
    end

    subgraph "Data"
        UDB[(User DB)]
        PDB[(Product DB)]
        ODB[(Order DB)]
    end

    MQ[Message Queue]

    Client --> Gateway
    Gateway --> US
    Gateway --> PS
    Gateway --> OS

    US --> UDB
    PS --> PDB
    OS --> ODB

    OS --> MQ
    MQ --> NS

Event-Driven Flow

sequenceDiagram
    participant UI as User Interface
    participant API as API Service
    participant Queue as Event Queue
    participant Process as Processing Service
    participant DB as Database
    participant Notify as Notification Service

    UI->>API: Create Order
    API->>DB: Save Order
    API->>Queue: Publish OrderCreated
    API-->>UI: 201 Created

    Queue->>Process: OrderCreated Event
    Process->>DB: Update Inventory
    Process->>Queue: Publish OrderProcessed

    Queue->>Notify: OrderProcessed Event
    Notify->>UI: Send Email/Push

Layered Architecture

graph TD
    subgraph "Presentation Layer"
        UI[Web UI]
        API[REST API]
    end

    subgraph "Business Logic Layer"
        Service[Services]
        Domain[Domain Models]
    end

    subgraph "Data Access Layer"
        Repo[Repositories]
        ORM[ORM/Query Builder]
    end

    subgraph "Infrastructure"
        DB[(Database)]
        Cache[(Cache)]
        Queue[Message Queue]
    end

    UI --> Service
    API --> Service
    Service --> Domain
    Service --> Repo
    Repo --> ORM
    ORM --> DB
    Service --> Cache
    Service --> Queue

Security Architecture Principles

Security Boundaries

1. Network Security

Internet → WAF/CDN → Load Balancer → Private Network
                                     ├─ Web Tier (DMZ)
                                     └─ App Tier (Private)
                                        └─ Data Tier (Isolated)

2. Authentication & Authorization

• Authentication: OAuth 2.0 + OIDC (Supabase Auth, Auth0, Cognito)
• Authorization: RBAC (Role-Based) or ABAC (Attribute-Based)
• API Security: JWT tokens, API keys with rate limiting

3. Data Protection

• In Transit: TLS 1.3
• At Rest: AES-256 encryption
• Secrets Management: Vault, AWS Secrets Manager, or environment variables

Security Checklist

• HTTPS everywhere (no HTTP)
• Input validation and sanitization
• OWASP Top 10 mitigations
• Rate limiting on public APIs
• SQL injection prevention (parameterized queries)
• XSS prevention (output encoding)
• CSRF protection
• Security headers (CSP, HSTS, etc.)
• Regular dependency updates
• Secrets never in code/repos

Scalability Patterns

Horizontal vs Vertical Scaling

Vertical Scaling (Scale Up):

• Add more resources to single server
• Simpler to implement
• Physical limits

Horizontal Scaling (Scale Out):

• Add more servers
• Load balancer required
• Better fault tolerance
• Preferred for high scale

Caching Strategy

Multi-Level Cache:

Request → CDN Cache → Application Cache → Database
          (Static)    (Redis/Memcached)

Cache Patterns:

1. Cache-Aside: App loads from cache, fetches from DB on miss
2. Write-Through: App writes to cache and DB simultaneously
3. Write-Behind: App writes to cache, async writes to DB

Database Scaling

Read Scaling:

• Read Replicas: Route read queries to replicas
• CQRS Pattern: Separate read and write databases

Write Scaling:

• Sharding: Partition data across multiple databases
• Federation: Split by function (users DB, orders DB, products DB)

Folder Structure Recommendations

Next.js App Router (2025)

src/
├── app/
│   ├── (auth)/               # Route group
│   │   ├── login/
│   │   └── register/
│   ├── (dashboard)/
│   │   ├── layout.tsx
│   │   ├── page.tsx
│   │   └── settings/
│   ├── api/
│   │   ├── users/route.ts
│   │   └── auth/route.ts
│   └── layout.tsx            # Root layout
├── components/
│   ├── ui/                   # shadcn components
│   ├── forms/
│   └── layouts/
├── lib/
│   ├── db/                   # Database client
│   ├── auth/                 # Auth utilities
│   └── utils.ts
├── hooks/                    # Custom React hooks
└── types/                    # TypeScript types

Python FastAPI

app/
├── api/
│   ├── __init__.py
│   ├── dependencies.py
│   └── routes/
│       ├── users.py
│       ├── products.py
│       └── orders.py
├── core/
│   ├── config.py
│   ├── security.py
│   └── database.py
├── models/                   # SQLAlchemy models
│   ├── user.py
│   └── product.py
├── schemas/                  # Pydantic schemas
│   ├── user.py
│   └── product.py
├── services/                 # Business logic
│   ├── user_service.py
│   └── order_service.py
├── main.py
└── tests/

Best Practices Summary

Separation of Concerns

Principle: Each component has single responsibility.

Example:

✅ Good:
- UserController: Handle HTTP requests
- UserService: Business logic
- UserRepository: Database operations

⌠Bad:
- UserController: HTTP + business logic + database

Dependency Inversion

Principle: Depend on abstractions, not concretions.

Example:

// ✅ Good
interface IEmailService {
  sendEmail(to: string, subject: string, body: string): Promise<void>;
}

class OrderService {
  constructor(private emailService: IEmailService) {}
}

// ⌠Bad
class OrderService {
  constructor(private emailService: GmailService) {} // Concrete dependency
}

API Design Anti-Patterns to Avoid

• ⌠Chatty APIs (too many requests needed)
• ⌠Overfetching (returning too much data)
• ⌠Deeply nested resources (/users/123/orders/456/items/789)
• ⌠RPC-style endpoints in REST (/getUserOrders, /createOrderForUser)
• ⌠Inconsistent naming
• ⌠Missing pagination
• ⌠No versioning strategy

Decision Framework Template

When recommending architecture, consider:

1. Current State:

- Existing system or greenfield? - Team skills and size? - Current pain points?

1. Requirements:

- Performance needs? - Scale expectations? - Budget constraints? - Time to market?

1. Trade-offs:

- Complexity vs flexibility? - Cost vs performance? - Speed vs correctness?

1. Future-Proofing:

- Growth trajectory? - Technology evolution? - Team expansion plans?

Output: Clear recommendation with reasoning, diagrams, and migration path (if refactoring).

Example Usage Patterns

Request: "Design architecture for a SaaS product with 100K users"

Response includes:

1. Recommended pattern (likely Modular Monolith → Microservices path)
2. Tech stack (Next.js + Supabase + Vercel)
3. System diagram (Mermaid)
4. Folder structure
5. Scaling plan
6. Security considerations
7. Migration strategy

Request: "Should I use REST or GraphQL for my e-commerce API?"

Response includes:

1. Comparison matrix
2. Recommendation based on context
3. Example schema/endpoints
4. Performance considerations
5. Implementation guidance

References

For detailed architectural patterns and examples, see:

• Official Next.js documentation for App Router patterns
• AWS Well-Architected Framework for cloud architecture
• Martin Fowler's patterns catalog for enterprise patterns
• Microservices.io for microservices patterns
• The Twelve-Factor App for modern app principles