ReleaseBytes
gcp Google Cloud Blog ·

Memorystore for Valkey 9.1 GA: Up to 3x QPS at Microsecond Latency

blogaigcpgaarchitectfinanceretailmedia
feature announcement

Google Cloud has announced the general availability of Memorystore for Valkey 9.1, offering up to three times higher queries per second and microsecond latency compared to Memorystore for Redis Cluster. This release is critical for organizations scaling AI and microservices, requiring massive throughput and low-latency data access. Key performance enhancements include a redesigned lock-free messaging architecture and dynamic thread scaling for improved I/O offloading. Valkey 9.1 also introduces granular database-level access control, the topology-aware CLUSTERSCAN command for efficient cluster-wide key scanning, and atomic operations like HGETDEL.

  • →Granular Database-Level Access Control (ACLs)
  • →CLUSTERSCAN for Efficient Cluster-Wide Key Scanning
  • →Atomic Fetch and Delete with HGETDEL
  • →Up to 3x QPS with New Thread Communication & Dynamic Scaling
Features (3) ›
  • Granular Database-Level Access Control (ACLs)

    Administrators can now restrict user access to specific numeric databases within the ACL framework, providing centralized management, secure multi-tenancy, and enhanced observability for enforcing least privilege. This foundational security mechanism is offered at no additional cost.

  • CLUSTERSCAN for Efficient Cluster-Wide Key Scanning

    The new topology-aware CLUSTERSCAN command allows reliable key scanning across large clusters, gracefully handling topology changes and redirections. It supports both sequential full cluster scans and parallelized scans for high-throughput requirements, preventing issues like missed keys or duplicates during slot migrations.

  • Atomic Fetch and Delete with HGETDEL

    Valkey 9.1 introduces the HGETDEL command, which atomically retrieves and deletes a hash field in a single network round-trip. This operation is particularly useful for application patterns involving single-use authentication tokens or short-lived session states.

Enhancements (1) ›
  • Up to 3x QPS with New Thread Communication & Dynamic Scaling

    Valkey 9.1 replaces static I/O thread assignment with a lock-free, multi-queue messaging architecture and a two-phase dynamic scaling engine. This eliminates cross-thread CPU waste, enables dynamic work balancing, and automatically adjusts I/O worker threads based on CPU usage and queue backlog for significantly improved performance.

Read the original announcement →

https://cloud.google.com/blog/products/databases/memorystore-for-valkey-9-1-3x-qps-caching/

Related releases