questdb/questdb

What it solves

QuestDB is a low-latency, open-source time-series database designed to handle massive ingestion rates (millions of events per second) while keeping years of historical data queryable. It eliminates the need for pre-aggregation and provides built-in tools for deduplication and out-of-order data correction.

How it works

The engine is built with zero-GC Java, utilizing C++ and Rust for performance-critical paths. Data is stored in memory-mapped, time-partitioned columns. It uses a tiered storage approach: a parallel write-ahead log, native columnar partitions, and Apache Parquet for long-term storage. Queries are executed across all CPU cores using SIMD and JIT-compiled filters.

Who it’s for

It is designed for industries requiring high-resolution telemetry and real-time analytics, including:

• Capital Markets: Tick data, order books, and post-trade analytics.
• Aerospace and Robotics: Flight-test telemetry and mission replay.
• Energy and Infrastructure: Grid and turbine telemetry.
• AI Coding Agents: Integration via MCP servers and specific agent skills.

Highlights

• High Performance: Capable of ingesting up to 19M rows/sec and streaming results to Apache Arrow at 220M rows/sec.
• Time-Series SQL: Extends standard SQL with specialized functions like SAMPLE BY, LATEST ON, and ASOF JOIN.
• QWP Protocol: A binary columnar protocol over WebSocket for high-efficiency reads and writes.
• Open Formats: Uses Apache Parquet, Iceberg, and Arrow to ensure data remains accessible to other tools.
• Broad Integration: Supports PostgreSQL Wire Protocol, InfluxDB Line Protocol, and REST APIs.