ChannelLife US - Industry insider news for technology resellers
United States
Valkey 9.2 cuts memory use with AI caching features

Valkey 9.2 cuts memory use with AI caching features

Thu, 19th Nov 2026 (Today)
Joseph Gabriel Lagonsin
JOSEPH GABRIEL LAGONSIN News Editor

Valkey has released version 9.2 of its open-source datastore, adding features aimed at reducing memory use and network traffic.

The release also includes a new distributed KV caching function for AI inference workloads. It creates a shared cache across GPU fleets so systems can reuse previous work instead of recomputing it.

Version 9.2 introduces forkless snapshots, continuous stream compression and a B+ tree architecture for large sorted sets. The changes target three common infrastructure costs at scale: spare memory reserved for backups, network traffic between regions, and memory overhead tied to large datasets.

Forkless snapshots are designed to remove the temporary memory spikes that have traditionally accompanied database backups on busy systems. Instead of using a separate process, the feature saves data through a background thread, allowing operators to avoid sizing systems for worst-case backup loads.

Another change addresses replication traffic across distributed deployments. Continuous stream compression cuts ongoing replication traffic by roughly 50% to 65%, according to Valkey, while backup file sizes can fall by up to 80% with the built-in LZ4 option. New replicas can also sync about 35% faster, with a CPU cost of about 2% to 4% of a core per replica.

For applications that rely on sorted sets, such as rankings, leaderboards and priority queues, the release rebuilds large sorted sets on a compact B+ tree structure. According to Valkey, this delivers up to 105% faster updates and 36% to 43% better memory efficiency for sets with millions of entries.

AI caching

Valkey is also introducing distributed KV caching for AI systems running open-weight models on self-managed inference infrastructure. The feature integrates with LMCache, which connects to inference engines including vLLM and SGLang through a single configuration change.

The goal is to reduce repeated processing of prompt instructions, tool definitions and conversation history in agentic AI applications. By sharing cache data across GPU fleets, users can cut AI inference costs by up to 40% and reduce time-to-first-token by up to 80%, according to the project.

The release also adds Path Hash, a native data type that acts as an index for AI caching workloads. It is designed to help users identify the longest matching prefix and locate cached results without building custom lookup logic, while also supporting other forms of hierarchical data.

"A major ethos of Valkey is doing more with less, which means getting more functionality with less hardware," said Madelyn Olson, Valkey Project Maintainer and Principal Engineer for AWS in-memory databases.

"We designed Valkey 9.2 around this principle to deliver reduced memory overhead, lower cross-region network costs and worry-free performance. With each release, Valkey continues to prove its value as an exceptionally fast, efficient and reliable datastore," Olson said.

Several companies involved in the Valkey ecosystem also highlighted contributions to the release and related projects.

"With representation on Valkey's Technical Steering Committee, Alibaba Cloud is committed to advancing Valkey through open collaboration. We are proud to have contributed Path Hash to Valkey 9.2, bringing native support for hierarchical data and expanding what developers can build with Valkey. Drawing on our Tair team's research and practical experience in KV cache for AI inference, we see significant opportunities for Valkey in AI infrastructure. We look forward to working with the community to help Valkey meet the evolving needs of AI workloads and deliver greater value to developers," said Guoyun Zhu, Head of Tair R&D at Alibaba Cloud.

Momento pointed to broader work on the software stack around the project. "Valkey has consistently delivered improved availability, performance and efficiency with each release. Valkey 9.2 is no exception, with forkless saves, hot key detection and even more efficient sorted sets. Momento and AWS have worked closely, and in the open, to advance RDMA for Valkey. It's a great example of the open-source ecosystem at work. Valkey proves that true open source does not lock innovation behind enterprise-license paywalls," said Khawaja Shams, Chief Executive Officer of Momento.

Multiverse Computing shared performance figures from its own use of distributed KV caching with Valkey. "At Multiverse Computing, we help customers scale inference workloads with compressed models that cut cost and latency. We cut both even further and served more prompts on existing GPUs by adding distributed KV caching on Valkey. With Valkey, we processed 50% more requests on the same GPUs within our one-second target for median time to first token. We achieved a 23% reduction in our median time to first token at normal traffic, and a 95% reduction at peak traffic. We also contributed LMCache connector improvements to the open-source community for Valkey that make distributed KV caching more resilient," said Ennio Pastore, Director of AI Platform at Multiverse Computing.