The short answer
Modern enterprise architectures demand real-time (CDC) from transactional operational databases without locking data into proprietary warehouse storage formats. Combining Google BigLake with Apache Iceberg provides an open, lakehouse architecture directly integrated into BigQuery.
The BigLake abstraction layer
Google BigLake extends BigQuery compute capabilities to open file formats stored on Google Cloud Storage, AWS S3, and Azure Blob Storage.
BigLake provides two foundational capabilities:
- Unified security governance: Fine-grained row-level, column-level, and dynamic data masking policies apply uniformly across open Parquet, ORC, and Iceberg files.
- BigLake Metastore: A serverless, fully managed Apache Iceberg and Hive-compatible metastore that coordinates transactional table metadata without requiring external catalog databases.
Continuous CDC ingestion into Iceberg
Incoming change streams (inserts, updates, and deletes) captured from operational databases by tools like or Apache Kafka can write directly to Cloud Storage as open Apache Iceberg v2 tables.
Because Iceberg supports ACID transactions with row-level positional delete files, streaming CDC ingestion updates table records continuously.
-- Query external Apache Iceberg table seamlessly using BigQuery SQL
SELECT customer_id, current_status, last_updated
FROM `my_project.lakehouse_curated.customer_iceberg`
WHERE current_status = 'SUSPENDED';Eliminating proprietary lakehouse lock-in
This architecture delivers distinct enterprise advantages:
- Zero data duplication: You do not copy data from Cloud Storage into BigQuery proprietary storage.
- Engine interoperability: Spark jobs running on Dataproc, Trino clusters, and external machine learning pipelines can read the exact same Iceberg tables simultaneously through the open Iceberg API.
- Accelerated analytics: BigQuery uses cached metadata and storage acceleration to deliver query performance comparable to native BigQuery tables while preserving complete data portability.