Horizontal Scaling

Overview

Horizontal scaling increases capacity by adding more service instances rather than increasing the size of a single machine.

How It Works

flowchart LR
    U[Users] --> L[Load Balancer]
    L --> I1[Instance 1]
    L --> I2[Instance 2]
    L --> I3[Instance 3]

Production Example

A traffic surge during a product launch is handled with autoscaling.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: web-api-hpa
spec:
  minReplicas: 3
  maxReplicas: 20
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 65

Design Requirements

• Stateless request handling
• Shared session store when session state is needed
• Consistent health checks and graceful startup
• Fast provisioning path for new instances

Common Pitfalls

• Scaling compute without scaling databases or caches
• Ignoring connection limits in dependencies
• Slow startup time causing delayed capacity response

Key Takeaways

• Horizontal scaling supports elasticity and fault tolerance.
• Stateless architecture simplifies scale-out strategies.
• Full-stack bottlenecks must be measured, not assumed.