SaaS · Cloud Engineering
Multi-region SaaS infrastructure
Active-active AWS architecture for a global trading SaaS with fault tolerance and latency-based routing.
- Availability target
- 99.999%
- Latency improvement
- ~350ms ↓
- Infra cost
- 18% ↓
- Regions
- 3 active
Problem
A late-stage FinTech SaaS ran a monolith in a single cloud region. Asian users saw multi-second lag vs New York; a regional outage became a business crisis.
They needed global availability, sub-100ms-class proximity routing, and infrastructure that could be rebuilt from code — not tribal knowledge.
Approach
Active-active across US-East, EU-Central, and AP-Southeast on EKS, with Route 53 latency-based routing.
Monolith decomposed into containerised services; Aurora Global Database for cross-region replication; Kinesis + Redis for real-time tickers.
Entire footprint in Terraform so recovery and environments were reproducible.
Outcome
True multi-region resilience with no single point of failure. Average global latency improved by ~350ms. Aggressive autoscaling and spot usage cut infrastructure spend ~18% while improving reliability.
Stack
- AWS EKS
- Aurora Global
- Terraform
- Node.js
- React
- Redis
- Kinesis