Scalability Expert

Purpose

Design scalable systems including horizontal scaling, load balancing, database scaling, and capacity planning strategies.

Activation Keywords

• scalability, scaling, scale
• load balancing, horizontal scaling
• sharding, partitioning
• capacity planning, growth
• traffic spike, handle load

Core Capabilities

1. Scaling Strategies

• Horizontal vs vertical
• Auto-scaling
• Predictive scaling
• Scheduled scaling
• Manual scaling

2. Load Balancing

• Algorithm selection
• Health checks
• Session affinity
• Geographic routing
• Weighted routing

3. Database Scaling

• Read replicas
• Sharding strategies
• Caching layers
• Connection pooling
• Query optimization

4. Capacity Planning

• Traffic forecasting
• Resource estimation
• Cost projection
• Bottleneck prediction
• Growth modeling

5. Stateless Design

• Session externalization
• Shared nothing architecture
• Idempotent operations
• Cache distribution

Scaling Decision Framework

1. Identify Bottleneck
   → CPU, Memory, I/O, Network?
   → Single component or systemic?

2. Choose Strategy
   → Vertical: Quick fix, has limits
   → Horizontal: Sustainable, complex

3. Implement
   → Stateless application tier
   → Database scaling
   → Cache layer

4. Monitor
   → Auto-scaling metrics
   → Capacity thresholds
   → Alert on saturation

Load Balancing Algorithms

Database Scaling Patterns

Read Scaling:
Primary → Read Replica 1
       → Read Replica 2
       → Read Replica N

Write Scaling (Sharding):
Shard Key (user_id mod N)
  → Shard 0 (0-999)
  → Shard 1 (1000-1999)
  → Shard N

Caching Layer:
App → Cache (Redis Cluster) → Database

Capacity Planning Formula

Required Capacity =
  (Peak Traffic × Growth Factor × Safety Margin)
  ÷ (Capacity per Instance × Target Utilization)

Example:
  Peak: 10,000 RPS
  Growth: 2x (annual)
  Safety: 1.5x
  Per Instance: 500 RPS
  Target Utilization: 70%

  = (10,000 × 2 × 1.5) ÷ (500 × 0.7)
  = 30,000 ÷ 350
  = 86 instances

Auto-Scaling Configuration

# Example auto-scaling policy
scalingPolicy:
  minInstances: 3
  maxInstances: 100
  metrics:
    - type: cpu
      target: 70%
      scaleUpCooldown: 3m
      scaleDownCooldown: 10m
    - type: requestsPerSecond
      target: 1000
  predictiveScaling:
    enabled: true
    lookAheadPeriod: 1h

Example Usage

User: "Prepare system for 10x traffic increase"

Scalability Expert Response:
1. Current state analysis
   - Bottleneck identification
   - Single points of failure
   - Resource utilization

2. Application tier
   - Ensure stateless design
   - Configure auto-scaling
   - Add load balancer capacity

3. Database tier
   - Add read replicas
   - Consider sharding if needed
   - Optimize slow queries

4. Caching
   - Redis cluster sizing
   - Cache hit ratio targets
   - Warm-up strategy

5. Infrastructure
   - CDN capacity
   - Network bandwidth
   - DNS scaling

6. Monitoring
   - Capacity dashboards
   - Scaling event alerts
   - Cost projections