Parallelism and Performance:
- DuckDB transitioned from pull-based to a push-based execution model to improve parallel processing.
- Multi-threaded execution and morsel-driven parallelism
- Optimized for multi-core systems
- DuckDB transitioned from pull-based to a push-based execution model to improve parallel processing.
- Multi-threaded execution and morsel-driven parallelism
- Optimized for multi-core systems
Comments
- Push-based execution architecture
- Centralized Control Flow
- Supports Asynchronous I/O
- Built-in backpressure handling
- Native Parquet reader/writer
- CSV and JSON support
- Direct connections to Postgres/SQLite
- Query external data sources
- ACID compliant
- snapshot isolation
- MVCC (Multi-version concurrency control)
- Optimized for vectorized processing
- Modular etension system
- WebAssembly support
- HTTP access capabilities
- Full-text search options
- JSON processing features