One interface for streaming & analytics
Replace pipeline sprawl with a unified, zero copy architecture powered by Kafka and Iceberg logical views. Ship real-time apps and durable analytics against the same tables.
<100ms
Fresh reads
0 copies
Single source of truth
Any engine
Iceberg native
Why combined views?
Most stacks evolved organically databases, warehouses, metrics, logs, BI, and ML stitched together by brittle jobs. The result is duplication, drift, and downtime. With combined views, Kafka is the source and Iceberg is the language. You get real-time and historical in one place, without the cost of copies.
Serve fresh operational data and durable analytics from one interface no laggy syncs or stale dashboards.
Keep a single source of truth in Kafka; expose it as Iceberg views without duplicating storage.
Schema evolution, retention, and lineage live in one place so policies don't drift across pipelines.
Logical views compute at query time, eliminating minutes to hours of batch lag.
Speed and batch both speak Iceberg mix and match without bespoke client code.
Move data in large, optimized chunks when it matters; avoid small file storms and compaction churn.
How it works
Logical views create an Iceberg facade over Kafka. Engines read views just like any other Iceberg table. When retention thresholds are reached, bulk moves graduate hot data from the speed layer to the batch layer efficiently and predictably.
On demand logical views transform Kafka topics into Iceberg compatible tables at query time.
Use your favorite engines (Trino, Spark, Dremio, DuckDB, Pinot, etc.) against the same tables.
When Kafka retention hits thresholds, bulk insert from speed → batch as native Iceberg operations.
Ready to try Combined Views?
Talk to our team about adopting logical Iceberg views over Kafka. We'll help you pilot in days and scale to production.
FAQs
answers to common questions about Combined Views and how it integrates with Kafka and Iceberg.
Connect style ETL creates a second, drifting copy and adds unavoidable latency. Logical views compute results at read time, so data stays single source and current.