Skip to content
BENIBIX Learn

Apache Iceberg vs Delta Lake

People think Iceberg and Delta Lake store data in different ways, and they do not. Both keep plain Parquet files and add a log of which files make up each version. Delta Lake began at Databricks and fits best there. Iceberg began at Netflix and is read by the widest mix of engines. Pick by the engines you run.

Data🗄️ Storage & Warehousing

Apache Iceberg

A table format that many tools share. It hides how the files are split.

Apache Iceberg is an open table format for huge tables in a data lake. It tracks which files make up the table, in layers of small files. Each change makes a new snapshot of the table. Many engines can use the same tables, such as Spark, Trino, Flink and Snowflake. It hides how files are split, so queries do not need to know. Old snapshots let you go back in time.

Example
A bank keeps one Iceberg orders table. Its analysts query it from Trino while a Spark job writes new rows.
Why it matters
It keeps data free of any one engine. You can change tools without moving the data.
Full entry
Data🗄️ Storage & Warehousing

Delta Lake

A table format first made by Databricks. It is Parquet files plus a log of changes.

Delta Lake is an open table format first built by Databricks. It stores data as Parquet files plus a folder called _delta_log. Each change adds a new entry to that log. The log gives safe, all-or-nothing writes, even with many writers. You can read the table as it was at any past version.

Example
A job deletes one customer's rows to meet a privacy request. Delta writes new files and logs the change as version 42.
Why it matters
It is the main table format on Databricks. So a large share of lakehouse data is kept in it.
Full entry

In one line

Apache Iceberg
A table format that many tools share. It hides how the files are split.
Delta Lake
A table format first made by Databricks. It is Parquet files plus a log of changes.