Dremio

Dremio runs queries as jobs: AddisDB submits the SQL, polls the job until it completes, then fetches the rows. That happens behind the scenes, so a long-running lakehouse query behaves exactly like a fast one in the editor.

Who it is for

Dremio queries data lakes directly — S3, ADLS, Iceberg, Delta — and adds a semantic layer of virtual datasets on top, so analysts get curated views without anyone copying data into a warehouse first.

Reflections are the feature that makes that practical: Dremio maintains materialized aggregations and sorted copies behind the scenes, and the optimizer silently rewrites a query to use one when it can. Your SQL names the virtual dataset; what actually gets read may be a reflection.

It suits organizations who want self-service BI straight on the lake, and teams trying to retire the "copy everything into a warehouse first" step.

Set up the server

1. Dremio Cloud: sign in and copy your project’s endpoint. Self-managed Dremio serves its API on port 9047.
2. Under your account settings, create a personal access token and copy it.
3. Add your data sources (S3, a database, a lake catalog) and note the space or path you will query.
4. Local: docker run -p 9047:9047 -p 31010:31010 dremio/dremio-oss, then complete the first-run setup.
5. Give the first boot a couple of minutes — the API answers before the catalog has finished coming up.

Connect from AddisDB

1. New Connection → Dremio under Big Data / Warehouse. Port prefills to 9047.
2. Enter the host, and paste your personal access token into the API key / token field.
3. Space is optional — leave it blank to browse everything.
4. For a different base URL (a proxy, or Dremio Cloud), paste the full URL into the Connection URL box.
5. Test, then Save.

Port 9047 is the REST API that AddisDB uses; 31010 is the Arrow Flight / ODBC port. Only the first one matters here.

Naming datasets

Paths are hierarchical and often contain characters SQL will not accept bare, so quoting each segment in double quotes is the habit that saves time. A physical dataset lives under its source; a virtual dataset lives in a space.

SELECT status, count(*) AS n
FROM   "Sales Analytics"."curated"."orders_v2"
WHERE  order_date >= DATE '2026-07-01'
GROUP  BY status
ORDER  BY n DESC;

What AddisDB gives you

• The Big Data Console — spaces and datasets in the catalog browser, a query pane, and a row cap on previews.
• Job submission, polling and result fetching handled for you.
• Full safety: read-only connections, prod tagging, destructive-statement detection.
• The Chart view, notebooks, ⌘K search, and AI querying grounded in your datasets.