Cloud-Native Change Data Capture: Real-Time Data Integration from Google Spanner to BigQuery

Abstract

This paper introduces a scalable, low-latency framework for real-time data replication from Google Cloud Spanner to BigQuery, leveraging Change Data Capture (CDC) via Google Cloud Dataflow. The proposed architecture captures transactional changes including inserts, updates, and deletions from Spanner using Change Streams and ensures their consistent delivery to BigQuery with minimal latency. Building upon Google's foundational CDC solution, the framework extends its capabilities through support for complex and nested data types, dynamic schema evolution, robust null value handling, and advanced exception processing. Key features such as checkpointing, dead-letter queue (DLQ) integration, and automated recovery enhance fault tolerance and system resiliency. This work also addresses limitations in Google's reference CDC implementation by introducing targeted enhancements for schema flexibility, fault resilience, and operational observability. The solution empowers enterprises to maintain continuously synchronized analytical datasets, enabling real-time insights and improved decision-making across business domains. Performance evaluations demonstrate high throughput, sub-second latency, and operational reliability, positioning the framework as a strong candidate for production-grade cloud-native data integration pipelines.