HTAP
ShannonBase builds upon the robust transactional foundation of MySQL, preserving full ACID compliance, row-level locking, and multi-version concurrency control (MVCC). By retaining these proven transactional capabilities, ShannonBase ensures reliable OLTP operations and strong consistency for mission-critical applications.
On top of this solid transactional core, ShannonBase introduces a high-performance in-memory columnar engine called Rapid, enabling Hybrid Transactional/Analytical Processing (HTAP). Transactional and analytical workloads are intelligently offloaded to either InnoDB or Rapid, using cost-based and machine-learning-driven scheduling, without sacrificing transactional integrity.
Additionally, ShannonBase extends its transactional engine with native AI capabilities. Users can perform machine learning training, prediction, and vector search directly within the database via SQL or JavaScript stored procedures. Changes in InnoDB are automatically and synchronously propagated to the HTAP engine, ensuring that analytical and AI workloads always work on consistent, up-to-date data.
By combining MySQL’s proven transactional reliability with advanced analytical and AI-native features, ShannonBase provides a single, unified platform for both OLTP and OLAP workloads, empowering developers and data scientists to build modern, AI-driven applications with confidence.
ShannonBase incorporates a columnar store, IMCS (In-Memory Column Store), named Rapid, to transform it into a MySQL HTAP (Hybrid Transactional/Analytical Processing) database. Transactional and analytical workloads are intelligently offloaded to either InnoDB or Rapid using a combination of cost-based and ML-based algorithms. Additionally, version linking is introduced in IMCS to support MVCC (Multi-Version Concurrency Control). Changes in InnoDB are automatically and synchronously propagated to Rapid by notification hook or applying Redo logs.
Rapid is ShannonBase's high-performance in-memory columnar engine, designed to accelerate analytical and hybrid workloads. By leveraging columnar storage, vectorized execution, and advanced compression, Rapid delivers sub-millisecond query response times on large datasets while maintaining transactional consistency through seamless integration with InnoDB.
With version linking and MVCC support, Rapid automatically tracks changes from the transactional engine, ensuring that both analytical and AI workloads operate on the most up-to-date data without requiring ETL pipelines or manual data synchronization.
Rapid natively supports complex data types, including structured, semi-structured, and vector embeddings, enabling machine learning inference, similarity search, and real-time analytics directly within the database. By combining these features, Rapid transforms ShannonBase into a true HTAP engine capable of serving both OLTP and OLAP workloads efficiently.
This architecture allows developers and data scientists to execute heavy analytical queries, AI model training, and prediction workloads in parallel with transactional processing, all while preserving consistency, reliability, and high performance.
