The lakehouse you actually own.
Your tables are Apache Iceberg, sitting in object storage you control. Four query engines read the same copy. Nothing is locked in a vendor format, and nothing has to be moved to be useful.
What "open" buys you
Every new Trino, Spark and Doris cluster writes Apache Iceberg as its table format. That is not a branding choice. It is the reason the rest of this page is possible:
- Schema evolution without rewrites. A column arrives, a type widens, the table keeps working.
- Time travel for a retention period you configure, so "what did this look like on Tuesday" is a query rather than a restore.
- Compatible reads across engines, with no translation layer in between.
Existing Delta Lake tables keep working. Databasin just stops creating new ones in that format.
Is this a data lake or a warehouse?
It is worth being precise, because the words get used loosely. A data lake is cheap storage holding files in open formats, and it will take anything you give it, which is also its problem: no schema guarantees, no transactions, and querying it well is somebody's full-time job. A warehouse gives you the guarantees but usually wants the data in its own format, on its own storage, at its own price.
A lakehouse keeps the cheap open storage of the data lake and adds the table guarantees of the warehouse: ACID transactions, schema enforcement and evolution, and time travel. Apache Iceberg is the part that makes those guarantees possible over plain files.
The practical difference for you: the storage account is yours, the format is readable by anything that speaks Iceberg, and if you stop paying us the tables are still sitting there.
Four engines, one copy of the data
The same SQL editor talks to several engines. You pick the one that fits the job and Databasin handles the plumbing. There is no separate extract for each.
| Engine | Best for | Shape |
|---|---|---|
| Apache Trino | Interactive SQL and federated queries across catalogs | Streams results, needs a cluster |
| Apache Doris | Real-time OLAP: low latency and high concurrency, so dashboards stay quick | Streams results, needs a cluster |
| Apache Spark | Heavy ETL, large-scale processing, advanced analytics | Batched, needs a cluster |
| DuckDB | Small and medium data, quick single-node work | Batched, single node |
| Databricks | Querying a Databricks workspace you already run | External, billed by Databricks |
Most teams land on Trino for the bulk of the work, Doris when latency and concurrency matter, and reach for the others on specific jobs.
Where does it run?
Two options, same platform. Databasin Cloud is fully managed and HIPAA-ready from the first day, and most teams start there because it takes about five minutes. Databasin Private installs the identical platform inside your own Azure tenant, on your network, under your policies, with your keys. Data never leaves.
Getting data into it
A lakehouse is only worth having once your systems are in it. Databasin ships 75+ connectors, and for the 30 native sources it builds the pipelines, maps the schemas and creates the gold views for you.