Google previewed a serverless Iceberg REST catalog at the Apache Iceberg Summit last month, enabling BigQuery and engines such as Spark, Flink and Trino to create, update and query the same Apache Iceberg tables without duplicating data. The change aims to simplify multi — engine lakehouses by removing the need for separate copies and reducing operational friction for teams that mix cloud — native analytics with open-table formats.
Technically, the preview extends BigQuery’s infrastructure to support Iceberg tables and brings a managed metadata plane to deployments that often rely on homegrown or external metadata services. The offering includes automatic table maintenance, transactional semantics, and built — in change — data replication to coordinate updates across engines that previously required manual pipelines and orchestration. The preview also introduces centralized table access controls so permissions can be applied consistently across different compute runtimes, addressing a frequent operational burden in multi — engine environments. By consolidating metadata and access policy enforcement, the catalog aims to lower governance overhead and make it easier to enforce consistent security and compliance rules across analytic workloads.
At Google’s Next ’26 event, the company expanded the interoperability story into a cross — cloud lakehouse posture, saying customers will be able to query Iceberg catalogs across AWS and Azure and interoperate with external platforms such as Databricks and Snowflake. In parallel, BigQuery ObjectRefs reached general availability, enabling combinations of structured Iceberg data with unstructured files in Cloud Storage for multimodal analysis and AI workflows. Product leads Yuriy Zhovtobryukh and Angela Soares emphasized keeping data in open formats while letting organizations run different processing and analytics tools on the same datasets. That approach is intended to preserve vendor neutrality and give teams flexibility to choose or switch engines without fragmenting the underlying data.
Practitioners say the integration could remove a practical “hidden tax” on Iceberg adoption. David Colbert pointed to real-world friction points — compaction, metadata management and orchestration — where choices about catalogs and control planes drive long-term optionality. The serverless catalog’s managed metadata, centralized controls and replication features are intended to reduce those frictions and simplify multi — engine ETL and analytic patterns.
Availability and next steps are clear: the serverless Iceberg REST catalog is in preview, Knowledge Catalog (formerly Dataplex) is offered in preview as a governance layer for metadata, lineage and access across systems, and BigQuery ObjectRefs are generally available. Builders evaluating lakehouse and AI workflows should test interoperability, transaction semantics and replication behavior across their preferred engines and cross — cloud setups before production rollout.
Sources
Replies (0)
No replies in this topic yet.
