Redis 7.4
What's new in Redis 7.4
Redis 7.4 introduces several new features and improvements aimed at enhancing memory efficiency, performance, and ease of use for various applications. These updates include support for hash field expiration, new memory-efficient data types for AI workloads, simplified secondary indexing, and time series optimizations. Additionally, Redis 7.4 brings several behavior and component changes. Below is a detailed breakdown of these updates.
New features
Hash field expiration support
Redis 7.4 adds the ability to set expiration times for individual hash fields or adjust their remaining TTL. This feature, long-requested by users, improves memory efficiency and performance, especially in caching and session storage scenarios.
New memory-efficient data types for AI workloads
With the growing demand for AI applications, Redis 7.4 introduces BFLOAT16 and FLOAT16 data types. These new types reduce memory usage by up to 47% and lower latency by as much as 59% under load, making them ideal for storing and processing vector embeddings in AI-powered applications, including vector databases and Retrieval Augmented Generation (RAG) systems.
Time series optimization with insertion filters
Redis 7.4 introduces insertion filters for time series data, allowing sensors to ignore new measurements when the differences in time or value are minimal. This feature helps reduce the size of time series data and boosts efficiency.
Improvements
Simplified secondary indexing
The Redis Query Engine now offers a more straightforward approach to secondary indexing with the addition of the TAG index type. Querying tags with special characters (like @ and .) is easier, as it no longer requires escaping; simply wrap query terms in double quotes. The update also includes improved handling of empty and missing fields, making the data model more flexible. Geospatial search has been enhanced with new operators, such as INTERSECT and DISJOIN, and memory usage reporting for indexes has been improved.
Changes
Behavior changes
Redis 7.4 includes behavior changes such as using jemalloc instead of libc for allocating Lua VM code. This adjustment reduces memory fragmentation and improves performance. Additionally, the ACL LOAD command has been modified to ensure that only clients with affected user configurations are disconnected, reducing unnecessary disruptions.
Component versions
The Redis version 7.4 includes the following components: