Enabling Multi-Region Data Systems at Notion

This post details the technical implementation of multi-region data systems at Notion to support data residency for EU customers. It covers the architectural changes required to isolate data processing and storage within specific regions, including regional data lakes powered by Debezium and Spark, regional AI infrastructure with vector databases and Kafka, a centralized job orchestration system (Airflow) that directs region-specific Spark jobs, and region-aware event logging using Cloudflare Workers and Snowpipe streaming connectors. Key patterns for maintainability such as a workspace-to-region mapping table and shared code helpers are also discussed.