Redis 7.0 Unveiled: New Functions, Multi‑Part AOF, Sharded Pub/Sub & Beyond

Core New Features Overview

Function replaces the traditional EVAL command family by allowing users to load custom function libraries that are automatically persisted and replicated, solving the long‑standing issue of Lua script loss during restarts or failover. FUNCTION LOAD stores the library in RDB/AOF and propagates it to replicas, while SCRIPT LOAD does not.

Multi‑part AOF introduces a base‑plus‑incremental file layout that eliminates the memory and I/O overhead of AOF rewrite by storing full snapshots and incremental changes in separate files. This design also enables point‑in‑time recovery and reduces resource consumption.

Sharded‑pubsub addresses the broadcast storm problem in Redis Cluster by partitioning channels across shards, so each node only processes the channels it owns, dramatically cutting unnecessary network traffic.

Client‑eviction adds a new maxmemory‑clients setting that caps the total memory used by client buffers and other non‑data memory, evicting the most memory‑hungry clients when the limit is exceeded, thereby protecting the database from out‑of‑memory failures.

Redis History Overview

Since its inception, Redis has released seven major versions. Highlights include cluster support in 3.0, lazyfree and PSYNC2 in 4.0, streams in 5.0, threaded I/O, TLS, and ACLs in 6.0, and the extensive feature set of 7.0.

Use‑Case Expansion and Outlook

The ecosystem now offers many modules (RedisSearch, RedisJSON, RedisGraph, RedisTimeSeries, RedisBloom) and commercial extensions such as Alibaba Cloud’s Tair, which provide richer data structures for multi‑model services. 2022 marked a surge in module adoption, pushing Redis from a pure cache toward a high‑performance, compute‑oriented in‑memory database.

Consistency and Persistence Comparison

Traditional Redis provides configurable persistence (RDB, AOF) and asynchronous replication, but lacks strong replica consistency. Commercial offerings differ: Tair Persistent Memory delivers optane‑based full persistence and semi‑synchronous replication; AWS MemoryDB offers physical log replication with strong consistency but incurs significant write‑performance penalties; open‑source Redis‑like systems built on RocksDB provide cheaper storage at the cost of higher operational complexity and weaker consistency guarantees.

Conclusion

Redis 7.0’s stability improvements, new function framework, and enhanced persistence mechanisms make it suitable for enterprise workloads. Continued community contributions, especially from Chinese developers, are vital for shaping future versions and ensuring that Redis remains a robust, open‑source foundation for modern data‑intensive applications.