Video explainers
Apache Iceberg, explained in under a minute each
Eight short, animated explainers covering the ideas that come up most often when teams adopt Apache Iceberg. There is no narration — every point is on screen, over a light music bed — so they read just as well muted, in a meeting, or embedded in your own docs.
-
01 What Is Apache Iceberg?
Why a folder of Parquet files is not a table, how Iceberg's metadata tree fixes that, and what it buys you.
- A directory of files is not a table
- Iceberg tracks every file explicitly
- Database guarantees on object storage
-
02 Apache Iceberg V3
Deletion vectors, row lineage, VARIANT and the rest of the V3 spec, and how to adopt it table by table.
- format-version is a table property
- Deletion vectors replace positional delete files
- VARIANT, spatial types and nanosecond timestamps
- Row lineage and table encryption
-
03 What Is Compaction?
The small-file problem, what a compaction job actually rewrites, and why sorting matters as much as size.
- Frequent writes make thousands of tiny files
- Rewrite data files into fewer, larger ones
- Sorting and delete cleanup come with it
-
04 What Is Merge-on-Read?
Copy-on-write versus merge-on-read, where the cost lands, and how V3 changes the maths.
- Copy-on-write rewrites the whole data file
- Merge-on-read records the change and moves on
- Fast writes, slower reads — until you compact
-
05 Catalogs & the Iceberg REST Spec
What an Iceberg catalog actually does, why the REST catalog spec exists, and who implements it.
- The catalog answers one question
- Every catalog needed its own client
- One open HTTP API any engine can speak
- Many implementations, one protocol
-
06 Why Interoperability Matters
What multi-engine access actually buys you, what every extra copy costs, and why an open spec is leverage.
- One table, many engines, no copies
- Every copy is a bill and a bug
- An open spec is an exit option
-
07 How Iceberg Reduces Cost
Where the savings in an open lakehouse actually come from — and which lever moves the needle most.
- Data lives in your object store
- Route each workload to the engine that fits
- The biggest lever is scanning less
- Fewer copies, less egress, less lock-in
-
08 Iceberg & Agentic AI
Context, governance, reproducibility and safe writes — what an Iceberg lakehouse gives an AI agent.
- An agent is only as good as the metadata it can read
- Governance belongs at the catalog, not the prompt
- Snapshots make an answer auditable
- Agents can write, safely
Click any video to play with sound. Background music only — no narration, so nothing is lost with the volume down.
📚 Go Deeper on Apache Iceberg
Alex Merced has authored three hands-on books covering Apache Iceberg, the Agentic Lakehouse, and modern data architecture. Pick up a copy to master the full ecosystem.