Apache Druid

Apache Druid is an OLAP database built for teams that need fast, concurrent analytics on massive datasets without pre-caching queries or pre-defining schemas. It targets use cases where query latency must stay low even as data volumes and concurrent users grow into the hundreds of thousands of queries per second.

Druid sits in the same space as tools like ClickHouse and QuestDB, but its architecture is distinctly oriented around streaming-first ingestion and elastic, loosely coupled components.

What makes it fast:

• Scatter/gather query engine preloads data into memory or local storage, avoiding data movement and network latency during query execution
• Automatic columnarization on ingestion, combined with dictionary encoding, bitmap indexing, and type-aware compression, means queries touch only the data they need
• Time-based indexing is a first-class citizen, which benefits time-series and event-driven workloads considerably

Streaming and batch ingestion:

Druid connects natively to Apache Kafka and Amazon Kinesis without additional connectors, supporting query-on-arrival at millions of events per second. Historical batch data and live streaming data are queryable through the same interface. Schema auto-discovery handles column detection and type inference automatically, updating as data evolves.

include configurable tiering with quality-of-service controls for mixed workloads, automatic continuous backup, multi-node replication, and automated recovery. These make it practical to run as a production system without constant manual intervention.