AWS Database Blog
Announcing Valkey 9.0 for Amazon ElastiCache
Amazon ElastiCache now supports Valkey 9.0. This brings the latest community-driven innovations from the Valkey open source project to address the performance and capability requirements of applications as they grow more data-intensive and latency-sensitive, such as real-time analytics, AI-driven retrieval, and high-throughput caching. For example, operating separate search infrastructure in addition to the cache adds cost and latency. High-volume pipelined workloads hit throughput ceilings that force over-provisioning. Managing the lifecycle of individual fields within a cached object requires workarounds that create key sprawl. Additionally, multi-tenant architectures in cluster mode demand complex key-prefixing schemes that complicate application code and migrations.
Valkey 9.0 addresses each of these directly with built-in full-text and hybrid search that can reduce or remove standalone search systems, engine-level optimizations that deliver up to 40 percent higher throughput for pipelined workloads, hash field expiration for fine-grained data lifecycle management, and multi-database support in cluster mode enabled deployments. In this post, we explore how these enhancements help customers build faster applications, streamline architectures, and support new real-time and AI-driven workloads.
Valkey recently marked its second anniversary with more than 100 million Docker pulls, broad adoption across nearly every major cloud provider, and accelerating ecosystem momentum. Valkey 9.0 continues that pace of innovation for AWS customers. Learn more in our separate Valkey Turns Two post. Valkey has quickly become a leading open source choice for developers seeking a high-performance, vendor-neutral in-memory datastore. Amazon ElastiCache customers have already adopted Valkey broadly to power caching, session stores, real-time analytics, queues, and increasingly AI applications. With Valkey 9.0, customers gain another wave of innovation focused on performance, functionality, and operational flexibility.
Build richer search experiences directly in your cache
Valkey 9.0 brings the latest search innovations from the valkey-search open source project to ElastiCache, providing real-time full-text search, semantic retrieval, filtering, and aggregations over terabytes of data with microsecond latency and throughput up to millions of requests per second directly inside your cache. Last year, we introduced vector search to Amazon ElastiCache for Valkey, so customers can achieve the lowest latency with the highest throughput at 95% recall rate among popular vector databases on AWS for semantic caching workloads. Valkey 9.0 builds on that foundation with full-text search, aggregation pipelines, and hybrid queries that combine text relevance with vector similarity across billions of embeddings from providers such as Amazon Bedrock, Amazon SageMaker AI, Anthropic, and OpenAI, with results reflecting completed writes.
This means you can support use cases such as product catalog search, document discovery, semantic retrieval, recommendation systems, anomaly detection and log analytics, conversational memory, and Retrieval Augmented Generation (RAG) architectures directly inside ElastiCache. By combining vectors, text search, and aggregations in one engine on continuously updated data, many workloads can reduce or eliminate the need to operate separate search infrastructure. To learn more, check out the Announcing aggregations on Amazon ElastiCache post and Full-Text, Exact-Match, and Range Search on Amazon ElastiCache post.
Up to 40 percent higher throughput using pipelining
Valkey 9.0 introduces engine-level optimizations that can deliver up to 40 percent higher throughput when using pipelining, including faster command parsing, improved memory prefetching, and more efficient processing of batched requests. By reducing CPU stalls and better utilizing modern processor architectures, pipelined workloads can process more operations per second with lower overhead. For applications that batch commands, such as high-volume APIs, event processing systems, gaming leaderboards, ad tech platforms, and microservices, this can translate into more throughput without adding nodes. Customers already using pipelining can benefit by upgrading to Valkey 9.0. To learn more about pipelining and getting started, refer to the documentationValkey pipelining page.
Hash field expiration
Hash field expiration adds the ability to apply TTLs to individual fields within a hash instead of expiring the entire key. This enables more precise lifecycle management for complex objects. Examples include expiring one-time verification codes while retaining a user profile, aging out individual session attributes, or automatically removing stale counters and metadata while preserving the rest of the record. This can reduce key sprawl and simplify application logic.
Run multi-tenant workloads in cluster mode enabled deployments
With numbered databases in cluster mode, Valkey 9.0 helps customers run multi-tenant workloads with stronger isolation and more efficient infrastructure. Numbered databases act as lightweight namespaces, allowing the same key names to exist in separate logical databases without key collisions or complex prefixing schemes. This can streamline multi-tenant workloads, such as SaaS architectures, where each tenant or environment can use its own database and reduce application complexity by avoiding tenant prefixes embedded in every key.
This feature also helps streamline migrations from standalone environments that already rely on multiple databases, reducing refactoring effort when moving to distributed architectures. Common use cases include separating development, test, and production workloads, isolating customer datasets, preserving legacy application assumptions, and staging data migrations or A/B testing scenarios on the same cluster.
Because data remains distributed across the cluster, customers can combine logical separation with the scale, performance, and availability benefits of cluster mode enabled ElastiCache deployments. To learn more, refer to the Valkey numbered databases post and ElastiCache documentation.
Polygon queries for geospatial indices
Valkey 9.0 also expands geospatial search with polygon queries for geospatial indices. In addition to radius and bounding-box searches, applications can now use GEOSEARCH with the BYPOLYGON option to find members located inside an arbitrary polygon, making it easier to model real-world boundaries such as delivery zones, service territories, neighborhoods, geofences, venues, campuses, and operational regions. This gives location-aware applications more precise filtering directly in Valkey, without approximating irregular areas with multiple radius or box queries or pushing that logic into a separate geospatial system.
Conclusion
Valkey 9.0 enhancements bring the performance, search capabilities, and operational flexibility needed to power increasingly demanding real-time and AI-driven workloads while simplifying the architectures behind them. Valkey 9.0 for Amazon ElastiCache is available today at no additional cost in all AWS Regions. To get started with Valkey 9.0 on ElastiCache, you can:
- Create your first Valkey 9.0 cache: See the ElastiCache Getting Started tutorial to launch a new cache in minutes.
- Upgrade an existing cluster: Follow the upgrade documentation to upgrade from any Redis OSS or Valkey version to Valkey 9.0 in minutes with zero downtime. With minimal behavioral changes, upgrading requires no application changes for most workloads.
- Move existing self-hosted workloads to ElastiCache: Use the ElastiCache Online Migration to migrate your data from your self-hosted open-source Valkey or Redis OSS on Amazon Elastic Compute Cloud (Amazon EC2) to Amazon ElastiCache. All the new capabilities introduced in Valkey 9.0 are drop-in compatible with Redis OSS, allowing migration to Valkey 9.0 on ElastiCache with no application changes for most workloads.
- Explore the new capabilities: To get started with the new capabilities introduced in Valkey 9.0 for ElastiCache, check out the documentation for Valkey Search, hash field expiration, and numbered databases.