Scaling Strategy

GospeLib's infrastructure evolves across three phases. Decisions in Phase 1 must not block Phase 2 or Phase 3. The code stays the same — only the deployment topology changes.

Phase Summary

graph LR
    P1["Phase 1: Solo/Seed<br/>0-10K users<br/>$150-300/mo"] --> P2["Phase 2: Growth<br/>10K-100K users<br/>$1.5-3K/mo"]
    P2 --> P3["Phase 3: Enterprise<br/>100K+ users<br/>$5-10K/mo"]

Phase 1: Solo/Seed (0–10K users)

Timeline: Months 0–18
Team: 1 engineer
Infrastructure: Single-node k3s (or Docker Compose on a $40/month VPS)

Topology

graph TB
    subgraph Single Node
        GW["Gateway"] --> Content["Content"]
        GW --> Auth["Auth"]
        GW --> Billing["Billing"]
        GW --> AI["AI"]
        GW --> Notif["Notifications"]
        Content --> FDB["FalkorDB"]
        Content --> Redis["Redis"]
        Content --> TS["Typesense"]
        Auth --> PG["PostgreSQL"]
        Billing --> PG
    end

Key Characteristics

• Gateway + Content may run as a single process with chi router delegating by path prefix
• Auth service is just Clerk webhooks
• Billing is a single webhook handler + DB
• The monorepo structure is identical — code is separated at the module level even if it runs in the same binary
• All nx affected CI runs take <5 minutes

Cost Estimate

Resource	Provider	Monthly Cost
K3s node (or VPS)	EC2 t3.micro / DigitalOcean	$0–40
PostgreSQL	RDS db.t3.micro (free tier)	$0
Redis cache	ElastiCache t3.micro (free tier)	$0
FalkorDB	Docker on same node	$0
Container registry	ECR (500 MB free)	$0
DNS	Route53	$0.50
Secrets	AWS Secrets Manager	~$2
Total		~$2.50–42/mo

Phase 2: Growth (10K–100K users)

Timeline: Months 18–48
Team: 3–12 engineers
Infrastructure: EKS cluster, RDS Multi-AZ, ElastiCache

Topology

graph TB
    subgraph EKS Cluster
        Ingress["Nginx Ingress<br/>(ALB)"]
        GW["Gateway<br/>3 pods (HPA)"]
        Content["Content<br/>2 pods (HPA)"]
        Auth["Auth<br/>2 pods"]
        Billing["Billing<br/>2 pods"]
        AI["AI<br/>2 pods"]
        Notif["Notifications<br/>2 pods"]
        FDB["FalkorDB<br/>(StatefulSet)"]
        TS["Typesense<br/>(StatefulSet)"]
    end

    RDS["RDS PostgreSQL<br/>(Multi-AZ)"]
    EC["ElastiCache Redis"]
    CDN["CloudFront CDN"]

    Ingress --> GW
    GW --> Content & Auth & Billing & AI & Notif
    Content --> FDB & TS
    Content & Auth & Billing & AI & Notif --> RDS & EC

Scaling Triggers

Metric	Threshold	Action
Gateway CPU	>70%	HPA scales pods
Content CPU	>60%	HPA scales pods
FalkorDB memory	>80%	Vertical scale to r6i.2xlarge
PostgreSQL connections	>80%	Add read replicas
Typesense queries/sec	Degraded latency	Move to dedicated node

Key Challenges

• FalkorDB does not horizontally scale — vertical scale + Redis cache layer in front of graph (already designed in Phase 1)
• Typesense: single node sufficient to 10M documents; cluster for redundancy only
• PostgreSQL: read replicas for study data queries; write primary for auth/billing

Cost Estimate

Resource	Provider	Monthly Cost
EKS cluster (3 t3.medium)	AWS EKS	~$200
Compute (pods)	EC2	~$300–600
PostgreSQL	RDS db.r6g.large (Multi-AZ)	~$350
Redis	ElastiCache r6g.large	~$200
FalkorDB	Dedicated node	~$150
CDN	CloudFront	~$50
Monitoring	Grafana stack	~$100
Total		~$1,500–3,000/mo

Phase 3: Enterprise (100K+ users)

Timeline: Month 48+
Team: 12+ engineers
Infrastructure: Multi-AZ EKS, Aurora, service mesh

Additions Over Phase 2

Addition	Purpose
Service mesh (Istio)	mTLS between services, traffic management
HashiCorp Vault	Dynamic DB credentials, PKI, audit trail
Aurora Serverless v2	Auto-scaling PostgreSQL
ElastiCache Cluster Mode	Redis high availability
FalkorDB Enterprise	Graph clustering + HA
Multi-region	Data residency, latency reduction
SIEM	Security audit trail
SOC 2 Type II	Compliance certification

What Does NOT Change

Even at enterprise scale, these remain constant:

• API contracts — versioned from day 1
• Service boundaries — drawn correctly from day 1
• Database schemas — migrate forward, never backward
• Frontend code — scales independently via CDN
• Ingest pipeline — runs as a job, not a server
• The codebase — same monorepo, same module structure

Related Pages

• Architecture Overview — The three-phase evolutionary design
• Technology Stack — Version pinning across phases
• CI/CD Pipeline — Deployment flow at each phase
• Secrets & Network Policy — Phase-specific secret management