🏎️ TimescaleDB v2.21—ingest millions of rows/second and faster columnstore UPSERTs and DELETEs
TimescaleDB v2.21 was released on July 8 and is now available to all developers on Tiger Cloud.
Highlighted features in TimescaleDB v2.21 include:
- High-scale ingestion performance (tech preview): introducing a new approach that compresses data directly into the columnstore during ingestion, demonstrating over 1.2M rows/second in tests with bursts over 50M rows/second. We are actively seeking design partners for this feature.
- Faster data updates (UPSERTs): columnstore UPSERTs are now 2.5x faster for heavily constrained tables, building on the 10x improvement from v2.20.
- Faster data deletion: DELETE operations on non-segmentby columns are 42x faster, reducing I/O and bloat.
- Reduced bloat after recompression: optimized recompression processes lead to less bloat and more efficient storage.
- Enhanced continuous aggregates:
- Concurrent refresh policies enable multiple continuous aggregates to update concurrently.
- Batched refreshes are now enabled by default for more efficient processing.
- Complete chunk management: full support for splitting columnstore chunks, complementing the existing merge capabilities.
For a comprehensive list of changes, refer to the TimescaleDB v2.21 release notes.