GCP – Introducing Valkey 8.0 on Memorystore: unmatched performance and fully open-source

Editor’s note: Ping Xie is a Valkey maintainer on the Valkey Technical Steering Committee (TSC)

Today, we’re thrilled to announce Valkey 8.0 on Memorystore in preview, making Google Cloud the first major cloud platform to offer Valkey 8.0 as a fully managed service. Building upon the launch of Memorystore for Valkey 7.2 in August 2024, this further solidifies Google Cloud’s commitment to open source, providing you with the latest and greatest features from the Valkey open-source ecosystem. 

Valkey 8.0 on Memorystore is a testament to our commitment to supporting customers such as Major League Baseball (MLB). As the most historic professional sports league, MLB uses Memorystore to power its real-time analytics, processing vast amounts of data to provide fans with insights and statistics during games.

“At MLB, we’re obsessed with delivering the best possible experience for our fans. Valkey’s truly open-source approach to caching is a game-changer, promising the performance and innovation we need to keep fans engaged and connected. We’re excited to be part of this community and look forward to Valkey’s continued innovation on Memorystore.” – Rob Engel, Vice President of Software Engineering, Major League Baseball

The Valkey 8.0 release

Earlier this year, after Redis Inc. changed the license of Redis OSS from the permissive BSD 3-Clause license to a restrictive Source Available License (RSAL), the open-source community rallied to create Valkey (1, 2, 3) — a fully open-source alternative under the BSD 3-clause license. In just a few months, the Valkey community released the open source Valkey 8.0 in GA, showcasing the power of open-source collaboration and unfettered innovation.

Memorystore for Valkey 8.0 delivers enhanced performance, improved reliability, and full compatibility with Redis OSS — all as a fully Google managed service.

Improvements to the Valkey performance benchmarks are thanks to newly introduced asynchronous I/O capabilities. The enhanced I/O threading system allows the main thread and I/O threads to operate concurrently, enabling parallel processing of commands and I/O operations, and maximizing throughput by reducing bottlenecks in handling incoming requests. Memorystore for Valkey 8.0 achieves up to a 2x Queries Per Second (QPS) at microsecond latency when compared to Memorystore for Redis Cluster, allowing applications to handle higher throughput with similarly sized clusters. This makes Valkey 8.0 a great choice for high-throughput, real-time applications that aim to provide highly responsive user experiences.

Along with the throughput gain, Valkey 8.0 includes other optimizations that further enhance the overall speed of the service:

The SUNION command is optimized for faster set union operations.

The SDIFF and ZUNIONSTORE commands have been refactored for improved execution times.

The DEL command avoids Redundant deletions for expired keys.

CLUSTER SLOTS responses are cached for better throughput and reduced latency in cluster operations.

CRC64 performance is improved for large data batches, which is crucial for RDB snapshot and slot migration scenarios.

Valkey 8.0 also brings key-memory efficiency improvements, allowing you to store more data without requiring changes to your application. Keys are now embedded directly into the main dictionary, reducing memory overhead while enhancing performance. Additionally, the new per-slot dictionary splits the main dictionary by slot, further reducing the memory overhead by 16 bytes per key-value pair without degrading performance.

Meanwhile, Valkey 8.0 has improved reliability thanks to several features developed by Google that were subsequently contributed to the project, significantly enhancing cluster resilience and availability: 

Automatic failover for empty shards helps ensure high availability even during the initial scaling stages, allowing new, slotless shards to fail over smoothly. 

Replicating slot migration states helps ensure that all CLUSTER SETSLOT commands are synchronized across replicas before execution on the primary, reducing the risk of data unavailability during failover events, and enabling new replicas to automatically inherit the correct state. 

Additionally, slot migration state recovery ensures that after a failover, the source and target nodes are updated automatically, maintaining accurate routing of requests to the correct primary without operator intervention. 

Thanks to these enhancements, Valkey 8.0 clusters are more resilient against failures during slot movement, giving customers peace of mind that their data remains available even during complex scaling operations.

Compatible with Redis OSS 7.2

Just like Valkey 7.2, Valkey 8.0 maintains full backwards compatibility with Redis OSS 7.2 APIs, allowing for a seamless migration from Redis. Popular Redis clients like Jedis, redis-py, node-redis, and go-redis are fully supported so that migrating workloads to Valkey doesn’t require modifications to application code.

This fusion of open-source flexibility and managed service reliability provides you with a balance of control and convenience, making Valkey a great destination for your Redis OSS workloads.

Get started with Valkey 8.0 on Memorystore today

We invite you to get started with Valkey 8.0 on Memorystore today and experience the above enhancements for yourself. With features such as zero-downtime scaling, high availability, and RDB snapshot and AOF logging based persistence, Memorystore’s Valkey 8.0 provides the performance, reliability, and scalability today’s high demanding workloads deserve.

Get started today by creating a fully managed Valkey Cluster through the Google Cloud console or gcloud, and join the growing community that is shaping the future of truly open-source data management.