What’s New in Valkey
summary
Valkey 8.1 replaces the core hash table to cut per-item overhead (example: small key/value overhead drops from ~112 bytes to ~64 bytes, roughly 20% memory reduction) and ships a new bloom filter data type, HNSW vector search in Valkey Bundle, LDAP auth, and observability commands. Demo runs a single Valkey instance holding ~9,600 product embeddings plus two user embeddings, uses an on‑box tiny LLM for cached personalized descriptions, and shows bloom filter false positives at tiny capacity and fast hybrid keyword+vector results.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:48First year and performance
- 4:33Hash table memory access
- 7:43Valkey 8.1 engine
- 8:47Valkey Bundle extensions
- 10:53Bloom filters for sets
- 15:07Demo: Bloom filters
- 18:09Vector similarity search
- 21:40Demo: E-commerce personalization
- 32:24Scaling and roadmap
- 35:14Community resources
related talks
49mvalkey-bundle: One stop shop for real-time applications
33mDiscover Valkey: The Path to Now
06mKeynote: Cut Database Costs, Fund Innovation with Valkey
27mAn Introduction to Messaging in Valkey
48mValkey clusters unboxed
44mCustomizing Valkey with Rust: Building High-Performance Modules
41mValkey 8 — Release of a Redis Fork Focused on Performance
44mValkey: Latency in Cloud Environments