Skip to main content

Scaling Architecture

How to scale Bellamy Book for high traffic.

Scaling Strategies​

Horizontal Scaling​

Add more instances:

  • API — stateless; scale behind a load balancer
  • Frontend / Admin — static assets; scale or use CDN
  • Workers — scale per type: GraphWorker, ScoringWorker, TrendingWorker, HashtagWorker, ElasticsearchSyncWorker, BlogAutoGenerationWorker, MediaProcessingWorker, WebSocketWorker, WebPushNotificationWorker, ChatWorker (each consumes Kafka or runs as a hosted service)

Vertical Scaling​

Increase resources:

  • CPU
  • Memory
  • Storage

Load Balancing​

API Load Balancing​

         ┌─────────────┐
         │ Load        │
         │ Balancer    │
         └──────┬──────┘
                │
    ┌───────────┼───────────┐
    │           │           │
┌───▼───┐  ┌───▼───┐  ┌───▼───┐
│ API 1 │  │ API 2 │  │ API 3 │
└───────┘  └───────┘  └───────┘

Load Balancing Algorithms​

  • Round Robin
  • Least Connections
  • IP Hash
  • Weighted Round Robin

Database Scaling​

Read Replicas​

Master (Write) ──┐
                 │
                 ├──→ Replica 1 (Read)
                 ├──→ Replica 2 (Read)
                 └──→ Replica 3 (Read)

Sharding​

Partition data across multiple databases:

  • User sharding
  • Post sharding
  • Geographic sharding

Caching Strategy​

Multi-Level Caching​

1. Browser Cache
   ↓
2. CDN Cache
   ↓
3. Application Cache (Redis)
   ↓
4. Database

Cache Invalidation​

  • Time-based expiration
  • Event-based invalidation
  • Manual invalidation

Worker Scaling​

Auto-Scaling Workers​

Based on queue length:

  • High queue → Scale up
  • Low queue → Scale down

Worker Types​

  • Stateless workers (easy to scale)
  • Stateful workers (require coordination)

Performance Optimization​

Database Optimization​

  • Indexes
  • Query optimization
  • Connection pooling

Application Optimization​

  • Code optimization
  • Caching
  • Async processing

Monitoring Scaling​

Metrics to Monitor​

  • Request rate
  • Response time
  • Error rate
  • Resource usage
  • Queue length

Auto-Scaling Rules​

minReplicas: 2
maxReplicas: 10
targetCPUUtilization: 70
targetMemoryUtilization: 80

Next Steps​