Redisson 4.7 Released: New RMaps Batch Operations, Vector Sets, Jitter Reconnect, and Java 21 Virtual Threads

Forum topic · 小凯 · 2026-08-26

Summary

Redisson 4.7, the latest release of the widely used Java distributed data grid (IMDG) and coordination framework for Redis and Valkey, introduces five major pillars: batch operations for RMaps to cut network round-trips, a high-performance RCircularBuffer/RRingBuffer with fast sliding-window reads, a Vector Set for high-dimensional vector storage and KNN search supporting RAG workloads, adaptive Full/Equal Jitter delay strategies that eliminate thundering-herd reconnect storms, and full support for Java 21 virtual threads plus RESP3 protocol negotiation with Valkey, Redis 8.0, and Dragonfly. The release also adds throughput-limiting parameters (maxCount/maxSize) to RStream and builds on Netty 4.2 with Zstandard/LZ4 compression. The article reviews Redisson's core distributed locking mechanisms—Watchdog automatic lease renewal (30s lease, 10s renewal) and the RedLock quorum consensus algorithm—and explains how LocalCache with Pub/Sub invalidation broadcast delivers microsecond reads with eventual consistency. An executive summary recommends the open-source edition for standard locking, queuing, and rate limiting, and Redisson PRO for multi-datacenter Active-Active replication and two-level caching. Academic references include Lamport's logical clocks and Gray & Lamport's consensus on transaction commit.

Redisson 4.7, the latest release of the leading Java distributed data grid (IMDG) and coordination framework for the Redis/Valkey ecosystem, has been officially released alongside the newest Redisson PRO architecture. This deep-dive (translated from a zhichai.net analysis) examines the release's architecture and its implications for high-concurrency, cloud-native systems.

Key Points

Redisson 4.7 is built on five pillars:

New RMaps batch operations — pipelined, batch-memory aggregation that removes the network RTT bottleneck when operating on large numbers of distributed Map objects.
RCircularBuffer & RRingBuffer — array-based ring memory structures supporting ultra-low-overhead readNewest() / readOldest() sliding-window reads.
Vector Set — high-dimensional vector storage with Top-K nearest-neighbor (KNN) search, enabling Java developers to build enterprise RAG knowledge bases directly on Redis/Valkey without a dedicated vector database.
Adaptive Jitter reconnect strategies — FullJitterDelay / EqualJitterDelay randomize exponential backoff (Sleep = random(0, min(M, base * 2^attempt))), spreading reconnect requests across the timeline and preventing thundering-herd avalanches when hundreds of thousands of instances reconnect after a failover.
Java 21+ virtual threads and Valkey support — optimized Netty Future async chains coexist with VirtualThread (M:N scheduling); carrier threads are released when waiting on locks or I/O, supporting hundreds of thousands of concurrent resident tasks. RESP3 auto-negotiation enables deep compatibility with Valkey, Redis 8.0, and low-latency Dragonfly.

Release matrix (overview)

| Area | Feature | Value | |---|---|---| | Distributed collections | RMaps batch ops | Up to ~80% fewer network round-trips | | Streams | RStream maxCount() / maxSize() | Backpressure protection against consumer OOM | | Local caching | LocalCache + invalidation broadcast | ~10x read performance, millisecond-level invalidation | | Runtime | Netty 4.2, Zstandard/LZ4 compression, virtual threads | Reported 3–5x throughput gains |

Distributed Locking and Coordination

Redisson's lock lifecycle management uses Lua scripts for atomicity and provides:

Watchdog auto-renewal: when no leaseTime is specified, a background task renews the lock every 10 seconds (lease = 30s, renewal at lease/3), eliminating premature lock expiry during long-running business logic.
RedLock: quorum-based acquisition across N independent Redis masters — a lock succeeds only if acquired on at least ⌊N/2⌋ + 1 nodes within the validity window, mitigating lock loss from async master-replica failover.
RRateLimiter: distributed token-bucket rate limiting with millisecond-level token refill, smoothing flash-sale traffic spikes.

Resilience and Throughput

Full Jitter reconnect: prevents fixed-interval reconnect storms during Redis master-slave failover that would overwhelm the new primary.
LocalCache with Pub/Sub invalidation: PRO edition two-level caching where data updates broadcast invalidation messages to all nodes, yielding microsecond local reads with cluster-wide eventual consistency.

Enterprise Selection Guidance

Redisson Open-Source 4.7: recommended for standard distributed locks, queues, and rate limiting — covering roughly 95% of distributed coordination scenarios.
Redisson PRO: recommended for multi-datacenter Active-Active replication and heap-based two-level caching with invalidation sync.

The article concludes that Redisson 4.7 has evolved from a "Redis Java client" into a multi-engine, full-scenario distributed coordination and in-memory data grid foundation for the cloud-native era, with early support for the Linux Foundation's Valkey providing a vendor-neutral path amid the Redis license fork.

References

1. Lamport, L. (1978). *Time, clocks, and the ordering of events in a distributed system*. Communications of the ACM, 21(7), 558-565. DOI: 10.1145/359545.359563 2. Gray, J., & Lamport, L. (2006). *Consensus on transaction commit*. ACM Transactions on Database Systems (TODS), 31(1), 133-160. DOI: 10.1145/1132863.1132867