Is Apache Iceberg really better than Delta Lake for multi-engine data interoperability?

All answers to this question.

The "Iceberg vs Delta" debate has changed recently with UniForm (Universal Format), but Iceberg still holds the edge for pure multi-engine interoperability. Iceberg’s "Hidden Partitioning" is a massive benefit because users don't need to know the physical layout of the data to write efficient queries; the engine handles it. This prevents the common "full table scan" mistakes that happen in Delta if someone forgets to add a partition filter. However, Delta Lake is still generally faster within the Databricks ecosystem. If you are a heavy Snowflake and Trino user, Iceberg is likely the better choice for 2024 and beyond because it was built from the ground up to be engine-agnostic without a single company controlling the roadmap.

   Answered 2025-08-20 by Deborah Lewis

Does the "Write" performance of Iceberg suffer compared to Delta Lake when dealing with high-frequency small file updates or streaming?

   Answered 2025-09-05 by Kevin Parker

• Kevin, Iceberg can struggle with "small file syndrome" if you don't run regular maintenance (compaction) jobs. Delta has "Auto-Optimize" which handles this better out of the box. However, Iceberg 1.5+ introduced improved delete-on-write strategies that have closed the gap significantly. For streaming, you just need to be more diligent about your commit frequency and snapshot expiration settings in Iceberg.

     Commented 2025-09-15 by Charles Harris

We chose Iceberg because it allowed us to migrate from Spark to StarRocks without rewriting a single DDL or losing our time-travel capabilities.

   Answered 2025-09-25 by Steven King

• That's exactly why Iceberg is winning in the "Open" space, Steven. The ability to switch your compute engine while keeping your storage layer intact is the ultimate anti-lock-in move.

     Commented 2025-10-02 by Sarah Jenkins