doris

Doris is a database built for one job: answering analytics questions fast, even while the data is still streaming in. If you run dashboards that thousands of users hit at once, or you're piling up logs and metrics and want to query them with plain SQL, this is the kind of engine that sits underneath. It's fully open source under Apache 2.0, with no paid tier and no "cloud edition" holding features back.

Running it yourself is a real commitment. Doris is a distributed system written in Java and C++, with separate frontend and backend nodes to deploy, tune, and keep alive. There's a Kubernetes operator, Docker images, and connectors for Kafka, Flink, and Spark, but you're still operating a cluster. The newer decoupled mode splits compute from storage so you can scale query power against S3-style object storage, which helps, though it's not something you stand up in an afternoon.

The payoff is sub-second queries on huge datasets from one SQL engine that handles structured data, full-text, JSON, and vector search together, and it reads Iceberg, Delta Lake, and Hudi tables directly. Solo or small team: probably overkill, reach for ClickHouse or DuckDB first. A mid-size team with real-time analytics needs and someone who owns the infra: this is a serious, free alternative to Snowflake or BigQuery. Large team: worth a hard look before you sign a six-figure warehouse contract.

The catch is the operational weight. Snowflake and BigQuery cost money precisely because you never think about nodes, upgrades, or a cluster falling over at 2am. Doris hands you that bill in engineer-hours instead of dollars. If nobody on the team wants to own a database cluster, the free price tag is a mirage.