Cloud Parts

Dremio

A SQL lakehouse engine built on Apache Arrow, with its own Iceberg catalog and “reflections” that speed up queries without anyone changing them.

What it is
Lakehouses,Query engines
Type
Commercial product
Made by
SAP (acquired Dremio in July 2026)
License
Proprietary (Dremio Cloud and Enterprise); the free Community Edition is Apache 2.0 (dremio-oss), last released as 26.0.5 in September 2025
Facts checked
6 October 2026
Website Source code

What it is

A Dremio cluster has a coordinator and one or more engines made of executors. The coordinator plans each SQL query with a cost-based optimizer, and the executors read Parquet, Iceberg and Delta Lake data straight from S3, Azure or Google Cloud Storage into Apache Arrow buffers and process it there, keeping hot data in a columnar cache on their local disks. The same query can also reach databases such as PostgreSQL, Snowflake or MongoDB, so you can join lake tables with live systems. Analysts build views on top, and Dremio can back any table or view with a reflection: a precomputed copy, raw or pre-aggregated, stored as an Iceberg table. Queries keep naming the original view; the optimizer rewrites them to use a reflection when that plan is cheaper, and reflections on Iceberg tables refresh themselves when the table changes. BI tools connect over JDBC, ODBC or Arrow Flight SQL.

Since version 26.0 (April 2025), Dremio ships Open Catalog, a built-in Iceberg catalog based on Apache Polaris, which Dremio co-created. It speaks the Iceberg REST protocol, so Spark, Flink or Kafka Connect can write tables that Dremio then serves. There are three editions. Dremio Cloud is fully managed on AWS and bills $0.20 per Dremio Compute Unit (DCU); the smallest engine uses 16 DCUs an hour ($3.20) and a Medium 128 ($25.60), counted per second with a 60-second minimum (prices checked October 6, 2026). Dremio Enterprise is self-managed on Kubernetes with consumption pricing set by sales. The free Community Edition lacks the catalog, access control and Autonomous Reflections, and its open-source code hasn’t been updated since September 2025. SAP completed its purchase of Dremio on July 6, 2026.

Why teams pick it

  • Reflections speed up dashboards without changing their SQL: the optimizer substitutes them, and Enterprise and Cloud can create and drop them automatically from query patterns
  • Open formats end to end: Iceberg tables, Parquet files and a Polaris-based REST catalog that other engines can read and write
  • Federation: one query can join Iceberg tables with PostgreSQL, Oracle, Snowflake, MongoDB and other live sources
  • Arrow-native from storage to client: Arrow Flight SQL and an ADBC-based Power BI connector return columns without row-by-row conversion

Watch out for

  • The open-source Community Edition is falling behind: dremio-oss stopped at 26.0.5 (September 2025) while Enterprise moved to 26.1.x, and it has no role-based access control, hosted catalog or built-in monitoring metrics
  • SAP bought Dremio in 2026; SAP has pledged to keep contributing to Iceberg, Polaris and Arrow, but its announcements don’t say how standalone Dremio will be sold in future
  • Dremio Cloud is AWS only for now (Azure is listed as coming soon); Enterprise pricing isn’t published
  • Reflections are extra copies: they use storage and engine time to refresh, and on Cloud that refresh time is billed in DCUs like queries
  • Autonomous Reflections only cover Iceberg tables, Parquet datasets and views on them, and are capped at 100 reflections, 10 new per day

Alternatives

What all lakehouses have in common →
Amazon S3 TablesAWSA bucket type for Apache Iceberg tables: S3 stores them, acts as their catalog, and compacts and cleans them up for you.Apache AmoroIncubatingOpen sourceAn incubating Apache “lakehouse management system”, started at NetEase as Arctic, that watches Iceberg tables and keeps them fast on its own: it compacts small files, rewrites deletes and expires old snapshots and files in the background.DatabricksCommercialA managed lakehouse platform built on Apache Spark and Delta Lake, with Unity Catalog for governance, on AWS, Azure or Google Cloud.SpiralCommercialA closed-source storage and query engine for multimodal AI data (video, images, sensor traces and tables) that keeps everything as Vortex files in your own object storage, versions every change and streams results to GPUs; access is by request.Tiger LakeBetaCommercialTiger Cloud’s built-in Iceberg connector, in public beta: it copies a Postgres table or TimescaleDB hypertable into Apache Iceberg, then keeps it current from the write-ahead log.

How it fits with other technologies

An Iceberg table, a federated view and a reflection that speeds it up · sql

-- an Iceberg table in Open Catalog, partitioned by day
CREATE TABLE lake.sales.orders (
  order_id BIGINT, customer_id BIGINT, country VARCHAR, amount DECIMAL(10,2), ts TIMESTAMP)
PARTITION BY (day(ts));

-- a view that joins it with a live PostgreSQL table; nothing is copied
CREATE VIEW analytics.revenue AS
SELECT o.country, c.segment, CAST(o.ts AS DATE) AS order_day, o.amount
FROM lake.sales.orders o
JOIN crm_pg.public.customers c ON c.id = o.customer_id;

-- a precomputed aggregate that Dremio maintains
ALTER VIEW analytics.revenue
CREATE AGGREGATE REFLECTION revenue_by_day USING
DIMENSIONS (country, segment, order_day) MEASURES (amount (SUM, COUNT))
LOCALSORT BY (order_day);

-- dashboards still query the view; the planner uses the reflection
SELECT country, SUM(amount) FROM analytics.revenue
WHERE order_day >= DATE '2026-09-01' GROUP BY country;

Community, Enterprise and Cloud#

The engine is the same. What changes is who runs it, which catalog and security features unlock, and how often it gets updated. From Dremio’s editions page, checked October 6, 2026.

CommunityEnterpriseCloud
PriceFreeConsumption-based, via sales$0.20 per DCU, pay as you go or commit
Runs onDocker or Kubernetes, you manage itKubernetes, you manage itDremio-managed on AWS
Version26.0 (last release September 2025)26.1.xManaged by Dremio, zero-downtime upgrades
Open Catalog (Polaris)NoYesYes
Iceberg REST, Unity and Snowflake Open Catalog sourcesNo (Glue and Hive Metastore only)YesYes
ReflectionsManual and liveManual, live, autonomous (off by default)Manual, live, autonomous (always on)
Role-based access control, row and column policiesNoYesYes
LDAP or OpenID loginNoYesYes
Engine auto-start and stopNoYesYes, plus replica scaling
High availabilityNoNoYes
  • On Cloud, Dremio-hosted storage is billed per terabyte-month on top of DCUs; bringing your own S3 bucket avoids that charge. Calls to Dremio’s built-in LLM are billed per token.

Close alternatives#

Trino

Apache 2.0

Open-source SQL engine with many connectors that streams data between workers in memory. Its Iceberg connector has materialized views you refresh with REFRESH MATERIALIZED VIEW, and the catalog comes from elsewhere (Glue, a REST catalog, Polaris).

Amazon Athena

AWS

Serverless Trino-based SQL over S3 with the Glue catalog. Nothing to run and $5 per TB scanned on demand, so repeated dashboard scans add up; Iceberg tables must live in the Glue catalog.

See it in a real stack

No architecture uses Dremio yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2015–2026

Tomer Shiran, who had led product at MapR, Microsoft and IBM, and Jacques Nadeau, who ran MapR’s distributed systems team and was vice president of Apache Drill, founded Dremio in Mountain View in 2015. Their goal, in the company’s words, was to let analysts get the data they need “from any data source, then accelerate analytical processing for BI tools, machine learning, data science, and SQL clients.”

Nadeau wanted a standard in-memory columnar format so systems could stop converting data between them. That idea became Apache Arrow, launched as an Apache top-level project in February 2016 with Nadeau as its vice president, and Dremio’s engine reads every source straight into Arrow buffers. Dremio published its source under Apache 2.0 in July 2017, and version 2.0 in April 2018 extended Data Reflections, its acceleration layer.

Dremio later recast itself as an Iceberg lakehouse platform. It co-created Apache Polaris, an open Iceberg catalog, and 26.0 in April 2025 shipped a catalog built on it along with Autonomous Reflections. SAP announced in May 2026 that it would buy Dremio to combine SAP and non-SAP data for AI, and completed the deal on July 6.

Why it’s built this way

Arrow buffers from end to end
Converting rows between formats wasted CPU in every hop. Reading each source into Arrow, processing in Arrow and returning Arrow over Flight removes those conversions.
Reflections the user never names
Users query tables and views, never the reflections. The optimizer compares plans with and without each reflection and picks the cheaper one, so acceleration can be added or dropped without changing dashboards.
Query data where it lives
Instead of loading everything into a warehouse first, Dremio connects to object stores and databases and joins them in one query, with views as a shared semantic layer.
An open Iceberg catalog
Dremio says it co-created Polaris “because organizations demanded an open alternative to proprietary catalogs like AWS Glue and Unity Catalog”, so other engines can share the same tables.

Timeline

  1. 2015

    Founded by Tomer Shiran and Jacques Nadeau in Mountain View

  2. 2016

    Apache Arrow launched in February, with Nadeau as vice president

  3. 2017

    Source code published under Apache 2.0 in July (release 1.0.8)

  4. 2018

    Dremio 2.0 extends Data Reflections

  5. 2025

    26.0 ships a Polaris-based catalog; last Community release in September

  6. 2026

    Polaris graduates in February; SAP completes its acquisition on July 6

Sources24 · facts checked 6 October 2026

Improve this page on GitHub