Apache Druid

Apache Druid is an open-source analytics database for querying streaming and batch data. It is designed for analytics queries in milliseconds on high-cardinality datasets containing billions to trillions of rows, with workloads described from hundreds to 100,000 queries per second. Native Apache Kafka and Amazon Kinesis integrations support low-latency ingestion and queries as data arrives. Druid organizes ingested data into compressed, indexed columns with time indexing, dictionary encoding, and bitmap indexes. Queries can use Druid SQL or JSON-over-HTTP native queries, and joins are available during ingestion or at query time. Its web console loads data, manages datasources and tasks, displays server status and segments, and runs queries. Extensions connect to services and formats including S3, HDFS, Azure, PostgreSQL, Avro, ORC, and Parquet. Druid is free, licensed under Apache License 2.0, and downloadable for self-hosting on Linux, macOS, or other Unix-like systems; Windows is not supported. The local quickstart requires at least 6 GiB of RAM and Java 17. Security features are disabled by default, so production deployments need TLS, authentication, and authorization configured.

Apache Druid is aimed at real-time and batch analytics, with query, ingestion, and web-console options. Production users need to configure security, and Windows is not supported.

Druid may suit teams building low-latency analytics for streaming or batch data, especially workloads involving high query concurrency or immediate visibility into incoming data. It is intended for self-hosted use on Linux, macOS, or other Unix-like systems.

Overview

Apache Druid is an open-source database for real-time analytics across streaming and batch data. It is best suited to teams building user-facing or exploratory applications that need fresh results under heavy query traffic. Its speed and scaling ambitions are compelling, but operating a distributed system and securing it are part of the job.

Druid is designed for low-latency OLAP: the project describes millisecond queries on high-cardinality datasets containing billions to trillions of rows, and workloads ranging from hundreds to 100,000 queries per second. Those figures describe intended capabilities, not a guarantee for every deployment. Druid is also a poor fit for full-text search over text logs, though it can ingest and analyze semi-structured data such as JSON.

Key features

Streaming and batch ingestion

Native Apache Kafka and Amazon Kinesis integrations support low-latency ingestion and query-on-arrival, with ingestion designed for millions of events per second and guaranteed consistency. Druid also handles batch data, making it useful when teams need one analytics database for live events and larger datasets. Source coverage includes streaming systems, object stores, databases, and files.

Storage and query design

Ingested data is automatically columnarized, time-indexed, dictionary-encoded, bitmap-indexed, and compressed. That combination is built for analytical scans and fast filtering, rather than full-text search. Users can query through Druid SQL or JSON-over-HTTP native queries; joins are supported both during ingestion and at query time.

Operations and integrations

Ingestion, query, and orchestration components are loosely coupled, with deep storage supporting scale-up and scale-out. Continuous backup, automated recovery, and multi-node replication address durability and availability, while the web console provides tools to load data, manage datasources and tasks, inspect server status and segments, and run queries. Core extensions connect Druid to systems and formats including S3, HDFS, Google Cloud Storage, Azure, Kafka, Kinesis, Avro, ORC, Parquet, MySQL, and PostgreSQL.

Pricing

Apache Druid: free. The open-source analytics database is downloadable for self-hosting under the Apache License, Version 2.0. There is no paid tier or managed service described in this plan; teams should account for their own deployment and infrastructure. The project points users to Slack and GitHub for community help and names Cloudera, Datumo, Deep.BI, Imply, and Rill Data as commercial support providers.

Platforms

Druid supports API and web access and can be self-hosted on Linux, Mac OS X, and other Unix-like systems, including cloud environments such as AWS, GCP, and Azure. Windows is not supported. The local quickstart requires at least 6 GiB of RAM and Java 17. Security features are disabled by default, so production deployments must configure TLS, authentication, and authorization; documented authenticator extensions include HTTP Basic, LDAP, and Kerberos.

Who it's for

Druid fits teams serving low-latency, high-concurrency queries, exposing analytics in user-facing applications, exploring data ad hoc, or needing immediate visibility into streaming events. It is less suitable for Windows deployments, teams seeking a managed option in the described plan, or workloads centered on full-text search over logs.

Pros and cons

• Pro: Streaming and batch support, plus Kafka and Kinesis integrations, make fresh-event analysis a core use rather than an add-on.
• Pro: Columnar storage and multiple indexes are designed to keep analytical queries responsive across very large, high-cardinality datasets.
• Pro: SQL, native queries, and a web console offer both application-facing and operational paths into the database.
• Con: Security is off by default, leaving production operators responsible for configuring TLS, authentication, and authorization.
• Con: Windows is unsupported, and even the local quickstart calls for 6 GiB of RAM and Java 17.
• Con: Its distributed architecture and self-hosted plan require teams to take on deployment and operations rather than relying on a packaged managed service.

Alternatives

For a managed Kafka-oriented option with a free Basic plan, autoscaling Kafka resources, and more than 80 managed connectors, consider Confluent Cloud. Teams wanting self-managed stream processing with a stated limit of 24GiB memory and 48GiB disk could choose Materialize; its Cloud Capacity plan starts at 1.50. Ververica Platform offers a free Community Edition, though a free license may be required and its terms may restrict use.

For a portable open-source programming model whose execution costs depend on the runner and infrastructure, consider Apache Beam. Apache Flink is another free Apache-licensed option for stream processing. For a self-hosted distributed OLAP datastore, compare Apache Pinot. Apache Samza is an open-source stream-processing framework. Apache Spark is a free distributed data analytics engine with download, PyPI, Maven Central, and Docker options.

Explore more Streaming Analytics Software, OLAP Databases, OLAP Software, Columnar Databases, and Database Software.

Verdict

Choose Apache Druid if you need self-hosted, high-concurrency analytics over streaming and batch data, and have the operational capacity to configure and run it securely. Its ingestion and query design target workloads where fresh answers matter; look elsewhere if you need Windows support, full-text log search, or a managed service.