SAN FRANCISCO — Dremio, the Agentic Lakehouse company, today highlighted its leadership across the Apache Iceberg ecosystem, including V3 support now available in Dremio Cloud, the election of Dremio engineer JB Onofre to the Apache Software Foundation board, and continued momentum behind Apache Polaris. A longstanding advocate for open-source collaboration and the elimination of vendor lock-in, Dremio has made foundational contributions to projects including Apache Arrow (co-creator and core contributor), Apache Iceberg (contributor and leading educator), and Apache Polaris (co-creator). Reinforcing this commitment, JB Onofre, who shepherded Polaris through incubation, has been elected to the Apache Software Foundation board.
Iceberg V3 is designed to support more diverse and complex data types, offer greater control over schema evolution, and deliver performance enhancements for large-scale, high-concurrency environments. Dremio’s V3 integration advances handling of semi-structured data, row-level changes, and schema evolution, with full support in Dremio Cloud, including the VARIANT data type for JSON, deletion vectors for faster CDC (change data capture), and improved schema evolution.
“The Iceberg lakehouse has become the default architecture for AI and analytics,” said Rahim Bhojani, CTO of Dremio. “Most platforms added Iceberg as a feature, but Dremio was built on it from the ground up. Capabilities like Autonomous Reflections, Iceberg Clustering, and now V3 compound on each other, delivering an Iceberg platform that’s both the fastest and the easiest to manage.”
Dremio continues to set the standard for Apache Iceberg with:
- Apache Iceberg V3 Support: Dremio delivers full read and write support for the latest Iceberg specification. Deletion vectors accelerate row-level operations for CDC and streaming workloads. The VARIANT type eliminates the schema-on-write bottleneck for semi-structured data. Row-level lineage provides built-in creation and update tracking for regulated industries with no additional tooling required.
- Arrow-Based SQL Engine for Iceberg: Dremio’s query engine was built natively on Apache Arrow, the open columnar standard Dremio co-created, making it uniquely suited for Iceberg workloads. It processes Iceberg and Parquet data in vectorized batches without conversion to a proprietary format, delivering fast, scalable analytics with no lock-in.
- Autonomous Reflections: Dremio eliminates the management overhead of running an Iceberg lakehouse. Autonomous Reflections observe query patterns and automatically creates, refreshes, and retires materializations, accelerating queries from seconds to sub-second with no code changes or manual tuning. Reflections’ incremental refresh keeps data fresh at low resource cost.
- Iceberg Clustering: uses Z-order to co-locate data across multiple columns simultaneously. ith two-level pruning that skips data at both the manifest and row-group level, it minimizes I/O by running continuously on petabyte-scale tables without full-table rewrites. Automatic table maintenance: compaction, snapshot expiration, and orphan file cleanup run on policy-based schedules with no manual intervention, keeping tables performant and storage costs in check. Enables engineers to focus on building data products, not maintaining tables.
- Open Catalog (Powered by Apache Polaris): Dremio co-founded Apache Polaris, the open Iceberg catalog standard now graduated to a top-level Apache project. Built on Polaris, Dremio’s Open Catalog provides an Iceberg catalog that supports full read and write from any REST-compatible engine, including Spark, Flink, Trino, and DuckDB, all sharing the same Iceberg tables. Governance, including RBAC, row-level filters, column masking, and just-in-time credential vending, is enforced consistently at the catalog layer regardless of which engine is querying. Every Dremio-managed table is accessible to any Iceberg-compatible engine.
- Ingestion and Transformation: Dremio supports the full range of DML operations on Iceberg tables using standard SQL. Continuous ingestion via CREATE PIPE, batch loads via COPY INTO, and dbt Core integration make Dremio a complete platform for building and maintaining Iceberg-native data pipelines.
