Redis 8.8: New array data structure, rate limiter, performance improvements

2 days ago

Redis 8.8 introduces a new general-purpose array data structure, offering fast index-addressable storage, dynamic sizing, sparse support, server-side aggregation, and search capabilities.
Performance improvements across data types include up to 68% throughput gain for pipelined MGET with I/O-threads, up to −83% for XREADGROUP in streams, and up to 60% faster persistence and replication.
A window counter rate limiter command (INCREX) simplifies rate limiting without Lua scripts, supporting bounds, expiration control, and partial acceptance (saturation).
Streams gain message NACKing (XNACK) with SILENT, FAIL, and FATAL modes for explicit message rejection, prioritizing redelivery and handling poison messages.
Hash subkey notifications allow clients to subscribe to field-level events (e.g., expiration, deletion) via new pub/sub channel types, enhancing real-time monitoring.
Time series commands now support multiple aggregators (e.g., MIN, MAX, FIRST, LAST) in a single query, reducing round trips for use cases like candlestick charts.
JSON numeric arrays can be explicitly stored as BF16, FP16, FP32, or FP64, optimizing memory and precision tradeoffs for vector embeddings and AI workloads.
Sorted set union/intersection operations add a COUNT aggregator, enabling scores based on set membership count or weighted sum for ranking and analytics.

Hasty Briefsbeta