Apache Druid

Overview

Apache Druid is an open-source analytics database built for low-latency analysis of streaming and batch data. It targets workloads that need quick answers over large datasets, including user-facing applications, interactive exploration, and systems that must make new events available promptly.

The project describes OLAP queries running in milliseconds over high-cardinality data sets with billions to trillions of rows. Druid is designed to serve use cases ranging from hundreds to 100,000 queries per second while maintaining consistent performance. Those figures describe the system's stated capabilities, not a guarantee for every deployment: actual results depend on the data and infrastructure.

It combines ingestion, query, and orchestration components that can scale independently, with deep storage supporting scale-up and scale-out deployments. The project lists Apache Druid 37.0.0 as its latest stable release, released May 8, 2026. Its software and documentation use the Apache License, Version 2.0.

Key features

Streaming ingestion and query-on-arrival

Native integrations for Apache Kafka and Amazon Kinesis support low-latency streaming ingestion and queries on newly arriving data. Druid is designed for ingestion at millions of events per second with guaranteed consistency. It can also work with batch data, giving teams a database for both continuously arriving events and larger analytical datasets.

Indexed, columnar storage

As data is ingested, Druid automatically converts it to columnar form, indexes it by time, dictionary-encodes values, builds bitmap indexes, and compresses it. These structures are intended to support fast analytical queries, including searches across data with many distinct values.

Querying and management

Users can query through Druid SQL or native JSON-over-HTTP requests. SQL is the listed query language, and Druid supports joins both during ingestion and at query time. Its web console can load data, manage datasources and ingestion tasks, show server status and segments, and run SQL or native queries.

Extensions, recovery, and security

Core extensions connect Druid to storage systems, metadata stores, data formats, and other services. Supported examples include S3, HDFS, Google Cloud Storage, Azure, Kafka, Kinesis, Avro, ORC, Parquet, MySQL, and PostgreSQL. Reliability features include continuous backup, automated recovery, and multi-node replication.

Security needs deliberate setup: Druid's security features are disabled by default, so production deployments must configure TLS, authentication, and authorization. Documented authenticator extensions include HTTP Basic authentication, LDAP, and Kerberos.

Pricing

Apache Druid is free and open source. The listed plan is 0.00 USD per free, described as an open-source analytics database downloadable for self-hosting. No paid plan price is listed. The project directs users to Slack and GitHub for community help and lists Cloudera, Datumo, Deep.BI, Imply, and Rill Data as commercial support providers.

Platforms

Druid is available for self-hosted deployment and is listed for Linux, macOS, web, and API use. The local quickstart supports Linux, Mac OS X, and other Unix-like operating systems; Windows is not supported. Running the quickstart locally requires at least 6 GiB of RAM and Java 17. Druid can run on commodity hardware in *NIX environments and is designed for cloud deployments on AWS, GCP, Azure, and other providers.

Who it's for

Druid is aimed at teams building applications or analytics workflows that depend on low-latency queries, high concurrency, prompt visibility into new data, or ad hoc exploration. It is particularly relevant when streaming data is central to the workload, while its batch-data support broadens its use beyond live event pipelines.

It is less suited to full-text search over text logs, which the project says is not a common Druid use case. It can, however, ingest and analyze semi-structured data such as JSON. Deployment also calls for operational care: the local quickstart has defined Java and memory requirements, and production use requires configuring security rather than relying on defaults.

Pros and cons

Pros

• Built for low-latency OLAP queries over very large datasets.
• Native Kafka and Kinesis integrations support streaming ingestion and query-on-arrival.
• Automatic columnarization, time indexing, encoding, bitmap indexing, and compression prepare ingested data for analysis.
• Offers SQL and native JSON-over-HTTP queries, plus a web console for data and task management.
• Free, open-source software with a broad set of core extensions and cloud deployment options.

Cons

• Windows is not supported by the local quickstart.
• The quickstart requires Java 17 and at least 6 GiB of RAM.
• Security features are off by default and require configuration for production.
• Full-text search over text logs is not a typical use case.

Alternatives

For other streaming-oriented systems, consider Apache Beam, Apache Storm, Feldera, Timeplus, RisingWave, Materialize, or Confluent Cloud. Apache Spark is another option to explore for data processing needs.

For browsing by category, see Streaming Analytics Software, OLAP Databases, OLAP Software, Columnar Databases, and Database Software.

Verdict

Apache Druid is a focused choice for analytics systems where fast queries, substantial concurrency, and timely access to streaming data matter. Its SQL and native query interfaces, web console, storage indexing, and integration extensions make it suitable for a range of ingestion and analysis setups. It is not a turnkey choice for every environment: operators need to meet its local runtime requirements, configure production security, and select another approach if full-text log search is the primary requirement.