Harder, Better, Faster, Stronger - Building Valkey-Timeseries
summary
Valkey‑Timeseries is a Rust module on Valkey claiming full RedisTimeSeries API compatibility while adding active expiration background GC, per‑key compaction filters, and dependent compactions (example: 1m, 1h, 1d buckets). It stores data in configurable compressed chunks, uses string interning for labels, an inverted index implemented with an adaptive radix tree whose leaves are roaring bitmaps, and parallelizes fetch+aggregate across cores. Writes batch by series/chunk with MADD merges; joins include as‑of joins with tolerance and reducers. Clustering exists but needs more work.
AI-written summary · about
chapters (13)
- 0:00Introduction to Valkey-Timeseries
- 2:16What is a time series database
- 3:22Basic commands in Valkey-Timeseries
- 6:40Downsampling in Valkey-Timeseries
- 9:27Performance and quality improvements
- 16:24Query filter enhancements
- 18:14Data structure and storage
- 20:52Optimization techniques
- 22:58Indexing and adaptive radix trees
- 27:24Parallelization in queries
- 32:19Batch ingestion and merging
- 33:50Downsampling implementation
- 34:53Q&A session
related talks
33mDiscover Valkey: The Path to Now
48mValkey clusters unboxed
36mWhat’s New in Valkey
41mValkey 8 — Release of a Redis Fork Focused on Performance
27mAn Introduction to Messaging in Valkey
10mThe unreasonable effectiveness of key value databases
44mValkey: Latency in Cloud Environments
28mRust and Scala, Sitting in a Tree….