What's New

Amazon ElastiCache now supports Valkey 9.0, the open-source, vendor-neutral alternative to Redis, bringing a significant set of new capabilities to real-time and AI-driven workloads. Key additions include built-in full-text and hybrid search (combining vector similarity with traditional text search), up to 40% higher throughput for pipelined workloads, hash field-level TTL expiration, and multi-database support in cluster mode. These enhancements are part of a broader release containing more than 100 improvements, available at no additional cost across all AWS commercial, GovCloud, and China regions.

How It Works

• Valkey 9.0 operates as a drop-in, in-memory data store engine within ElastiCache, available on both node-based clusters and Serverless caches.
• The new full-text and hybrid search capability is built directly into the engine, enabling real-time indexing and querying—including semantic vector retrieval, filtering, and aggregations—over terabytes of in-memory data with microsecond latency and throughput scaling to millions of requests per second, eliminating the need for a separate search tier.
• The 40% throughput improvement for pipelined workloads is achieved through engine-level optimizations such as faster command parsing and improved memory prefetching, reducing CPU overhead per operation.
• Hash field expiration introduces per-field TTLs within a Redis/Valkey hash data structure, allowing individual fields to expire independently rather than requiring the entire key to be expired or managed externally.
• Multi-database support in cluster mode provides lightweight logical namespaces (analogous to the SELECT database feature in standalone Redis), enabling tenant isolation or environment separation without deploying separate clusters.

Why It's Important

• These capabilities directly address three common architectural pain points for teams building high-throughput, data-intensive applications.
• First, integrating full-text and hybrid search into the cache engine eliminates the operational overhead, cost, and latency of maintaining a separate search service (e.g., OpenSearch or Elasticsearch) alongside ElastiCache, which is particularly valuable for AI-driven retrieval-augmented generation (RAG) pipelines that require both vector and keyword search.
• Second, the 40% throughput gain for pipelined workloads means teams can handle higher request volumes on existing infrastructure, reducing the need to over-provision nodes and lowering costs.
• Third, hash field expiration and multi-database cluster support simplify data lifecycle management and multi-tenant architectures, reducing application-layer complexity and the workarounds previously required to achieve these patterns.

How It's Different

• Prior to Valkey 9.0, ElastiCache users who needed full-text or hybrid search alongside their cache had to integrate a separate service such as Amazon OpenSearch Service, adding latency, operational complexity, and cost.
• Valkey 9.0 consolidates these capabilities into a single in-memory engine.
• Compared to Valkey 8.x and earlier Redis-compatible versions on ElastiCache, the new release adds native hash field-level TTLs—previously, TTL granularity was limited to the entire key, forcing developers to use workarounds like storing fields as separate keys or managing expiration in application logic.
• Multi-database support in cluster mode is also a meaningful departure from prior cluster-mode behavior, which lacked logical namespace separation and made migrating from standalone Redis (which supports multiple databases natively) more complex.
• The 40% pipelined throughput improvement represents a meaningful engine-level advancement over prior Valkey/Redis versions without requiring any application changes.

When to Prefer It

• Valkey 9.0 on ElastiCache is the preferred choice when building applications that require low-latency, high-throughput data access combined with search capabilities—particularly AI and machine learning applications using RAG patterns that need both semantic vector search and full-text retrieval from the same data store.
• It is well-suited for workloads that rely heavily on command pipelining and are currently hitting throughput ceilings, as the engine-level optimizations provide a free performance uplift.
• Teams managing multi-tenant SaaS applications or migrating from standalone Redis environments with multiple logical databases will benefit from multi-database cluster mode support.
• Hash field expiration is ideal for use cases like session management, feature flags, rate limiting, or any scenario where individual attributes within a record have independent lifetimes.
• If your workload is already well-served by a dedicated search service with complex query requirements beyond what an in-memory store provides, or if you require durable, disk-based storage, a purpose-built service like OpenSearch may still be more appropriate.

Availability

• Valkey 9.0 for Amazon ElastiCache is generally available (GA) as of May 5, 2026, with no additional cost beyond standard ElastiCache pricing.
• It is supported on both node-based clusters and Serverless caches.
• The release is available in all AWS commercial regions, AWS GovCloud (US) regions, and AWS China regions.
• Customers can adopt Valkey 9.0 by creating a new cluster or upgrading an existing cluster via the AWS Management Console, AWS SDK, or AWS CLI.
• No specific instance type restrictions are called out in the announcement, though customers should consult the ElastiCache documentation for any node family compatibility requirements when upgrading existing clusters.