Scalable deployment patterns for BizFirstAI platform on cloud providers.
Providers: AWS, Azure, GCP
Options: Docker, Kubernetes
API Container
Base: mcr.microsoft.com/dotnet/aspnet:9.0
Size: ~200MB
Tags: latest, v1.0.0, rc-*
Frontend Container
Base: nginx:latest
Size: ~50MB
Content: Vite optimized build
Registry
Azure Container Registry
or Docker Hub / ECR
Namespaces
production, staging, development
Deployments
API (3-10 replicas)
SignalR Hub (2-5 replicas)
Jobs (1-3 instances)
Services
ClusterIP (internal)
LoadBalancer (external)
Scale up when CPU > 70% or Memory > 80%
Scale down when CPU < 30% for 5+ minutes
Round-robin across API instances
Sticky sessions for SignalR
Redis backplane for SignalR
Enables cross-instance communication
Route read queries to replicas
Reduce load on primary
Redis for session + query cache
TTL: 15 mins for session, 5 mins for queries
Min: 10, Max: 100 connections
EF Core manages automatically
| Scenario | Concurrent Workflows | API Instances | Database Tier | Est. Monthly Cost |
|---|---|---|---|---|
| Small (Dev/Test) | 10-50 | 1-2 | Single instance (shared) | $100-200 |
| Medium (Production) | 100-500 | 3-5 | Standard (HA) | $500-1,500 |
| Large (Enterprise) | 1,000-10,000 | 10-20 | Premium (Multi-region) | $2,000-5,000 |