Havenask is a large-scale distributed information search system developed by Alibaba Group, designed to provide high-performance, low-cost, and user-friendly search services for businesses. It supports real-time search on hundreds of billions of data records, achieving millions of queries per second (QPS) and writes per second (TPS) with millisecond query latency and second-level data updates.

How It Works

Built on a C++ core for enhanced performance, memory efficiency, and stability, Havenask offers a familiar SQL query interface. Its architecture emphasizes flexibility through a rich plugin mechanism for scalability and graphical development capabilities that allow for rapid algorithm iteration, making it suitable for next-generation intelligent search applications.

Quick Start & Requirements

• Install/Run: Use provided Docker images (ha3_runtime for running, ha3_dev for development). A create_container.sh script is available.
• Prerequisites: Docker installed and running. For single-node mode, passwordless SSH to localhost is required.
• Resource Requirements:
  • Runtime: CPU > 2 cores, Memory > 4GB, Disk > 20GB.
  • Development: CPU > 2 cores, Memory > 10GB, Disk > 50GB.
• Documentation: Havenask Wiki

Highlighted Details

• Supports real-time search on hundreds of billions of data records with millions of QPS/TPS.
• C++ underlying structure for performance, memory, and stability.
• SQL query support for a user-friendly experience.
• Rich plugin system for scalability and customization.
• Supports graphical development for rapid algorithm iteration.
• Supports vector search for multimodal search scenarios.

Maintenance & Community

• Contact via DingTalk group (details in README).

Licensing & Compatibility

• License information is not explicitly stated in the README.

Limitations & Caveats

• The README notes significant differences between old and main branches, advising users to ensure code consistency with the current main branch. Docker images are currently only available for amd64 architecture.