Dremio
Dremio is a SQL engine for the lakehouse that federates queries across Iceberg, Delta Lake, databases, and warehouses, and accelerates them with materialised structures it calls Reflections — available as a free Community edition, a self-managed Enterprise edition, and a managed cloud service.
LicenseOpen source (permissive)· Apache-2.0
DeploymentSelf-hostedManaged
PricingFreeSubscription
Operational complexityMedium
WorkloadInteractiveBatch
LanguageJava
Use it when
- BI dashboards hit lake data directly and need acceleration nobody manages by hand; Reflections rewrite queries onto materialized structures automatically.
- You want a semantic layer and governance over Iceberg and Delta without copying data into a warehouse first.
- Analysts join lake tables against operational databases and warehouses in one query.
- You want a genuinely free self-managed edition of the same engine before committing to a paid tier.
Think twice when
- Row-level policies and column masking matter and you plan to stay on Community; those are Enterprise and Cloud features.
- Delta Lake is your write format; Dremio writes Iceberg but reads Delta only.
- You need the widest connector surface; Trino and Starburst reach more systems.
- A failed long query reruns from the start; there is no mid-query fault tolerance.
How it runs
Community runs on Docker or Kubernetes anywhere; Enterprise is self-managed on Kubernetes; Cloud is fully managed on AWS and Azure with consumption billing. Multiple engines autoscale with routing rules per workload. Clients connect over Arrow Flight SQL first, with JDBC, ODBC, REST, and BI connectors on top.
Details
CompareHow Dremio answers the questions Query Engines turns on.
Query Engines
| How it works | |
| Engine shape | Distributed cluster; Cloud is fully managed |
| Catalog required | Connects to Glue, Unity Catalog, Snowflake Open Catalog, Google Cloud Lakehouse Catalog or any Iceberg REST catalog; its own Open Catalog needs nothing stood up |
| SQL dialect | Dremio SQL (ANSI-based) |
| Running it | |
| Concurrent users | Multiple engines with a concurrency limit each, routing rules by user, group, job type or tag, and autoscaling between minimum and maximum replicas |
| Acceleration | Reflections; autonomous ones in Enterprise and Cloud build themselves from a daily pass over seven days of query traffic, up to 100 with 10 new a day, on Iceberg and Parquet, Community gets live reflections only |
| Failure recovery | Nothing like spooled exchanges: a failed query is rerun; engine replicas cover availability rather than mid-query recovery |
| Connections | |
| Non-lake sources | A curated list rather than a long one, Postgres, MySQL, Oracle, SQL Server, Db2, Snowflake, Redshift, BigQuery, Synapse, SAP HANA, Vertica, MongoDB and Druid, plus S3 and Azure Storage |
| Table formats | Parquet, JSON and delimited text; Iceberg has full DML with time travel and partition evolution, Delta Lake is read-only |
| Client access | Arrow Flight SQL first (the JDBC and ODBC drivers are built on it) plus a REST API and connectors for the usual BI tools |
| Access | |
| Access control | Role-based privileges plus row-access and column-masking policies written as functions, an Enterprise and Cloud feature, not in Community |
Alternatives to Dremio
