Key takeaways from the first in-person Apache Iceberg Summit 2025 in San Francisco. Ryan Blue's keynote covered Table Spec V3 features (VARIANT type, materialized views, binary delete vectors, row lineage) and a V4 vision addressing small-file commit overhead. Sessions explored non-Java implementations: PyIceberg uses Rust bindings for performance while Go targets lightweight tooling and direct integrations. A talk on the 'deconstructed database' framed Iceberg alongside Substrait, Parquet, and Arrow as modular components of modern analytics infrastructure. Despite gaps in real-time and small-table support, Iceberg is considered to have crossed the adoption chasm as the dominant open table format.

4m read timeFrom quesma.com
Post cover image
Table of contents
Keynote: “v3 and Beyond: Iceberg’s Ongoing Evolution”Iceberg Beyond Java: A Layered ApproachThe Deconstructed DatabaseSummary

Sort: