Skip to main content
Version: v0.1.0

Scaling Strategies

Guide to handling growth in users and data.

1. Application Layer (Stateless)​

Horizontal Pod Autoscaling (HPA)​

The API and Frontend layers are stateless. We use K8s HPA to scale pods based on CPU/Memory/Request count.

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: ciyex-api
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: ciyex-api
  minReplicas: 3
  maxReplicas: 50
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70

2. Database Layer (Stateful)​

Read Replicas​

Offload heavy read operations (Reports, Dashboards) to Read Replicas. Spring Boot routes @Transactional(readOnly=true) transactions to replicas.

Partitioning​

Large tables (Audit Logs, Observations) are partitioned by date or Tenant ID.

Connection Pooling​

Use PgBouncer ahead of Postgres to multiplex connections from hundreds of pods.

3. Caching Layer​

  • Redis Cluster: Scale Redis horizontally for session and data caching.
  • CDN (CloudFront): Offload static assets (JS, CSS, Images, Docs) to the edge.

4. Multi-Tenancy Scaling​

Database-per-Tenant vs Schema-per-Tenant​

For massive organizations (Enterprise), we can move them to a dedicated Database Shard to isolate load.

Load Testing Triggers​

Trigger scaling events when:

  • CPU > 70%
  • Connection Pool usage > 80%
  • API Latency p95 > 500ms