Multi-Country GCP Infrastructure Architecture

Engineering challenge

Moving toward a multi-country service model required clearer boundaries for compute, databases, networking, and shared services. The target design also needed to account for dependencies already present in development environments.

Architecture & design

• Keep the current production architecture (As-Is) distinct from the target design.
• Separate shared VPC and services from regional resources and country-specific configuration.
• Mark databases, cache, and messaging as existing or simulated rather than assuming they are already isolated.
• Align BigQuery and Datastream ownership with the broader regional scope model.

Implementation

• Inventoried compute, PostgreSQL, Redis, Nginx, webhook proxy, GCS, Firestore, BigQuery, and MQTT dependencies.
• Planned GKE deployment boundaries and Terraform environment structure for country-level services.
• Updated overall and per-environment architecture diagrams to show component dependencies.
• Adjusted Test, Auto-Test, and regional Terraform configuration as the shared-resource model evolved.

Validation & impact

• Architecture diagram structure checks passed, and selected environment and Terraform configuration changes were recorded.
• Separating current and target states made shared dependencies and incomplete isolation items visible for review.

Limitations & next steps

• Deployment of all country environments is not confirmed.
• Cross-region traffic failover, disaster recovery, and data-policy validation are not yet evidenced.