How to Build AI-Scale Caching with Memorystore for Valkey
Read the Portuguese translation →
Author: Google Cloud · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: How to Build AI-Scale Caching with Memorystore for Valkey · Channel: Google Cloud · 03:15 · Watch on YouTube
Topics: Enterprise AI applications, Data, databases and embeddings, AI infrastructure and hardware
Summary
Section titled “Summary”Memorystore for Valkey is presented as a fully managed, open-source in-memory database service that provides sub-millisecond access, scalability, and high availability. It can sit between applications and backend databases, and offers compatibility as a drop-in replacement for Redis OSS workloads. The video highlights performance improvements, advanced caching and vector search capabilities, and migration tools designed to avoid downtime. Google Cloud manages operations such as failover, patching, scaling, and security, with enterprise features and four nines availability.
Key points
Section titled “Key points”- Memorystore for Valkey provides sub-millisecond data access as a fully managed in-memory database service.
- It can be deployed as a cache between applications and backend databases such as Cloud SQL, AlloyDB, Spanner, and Firestore.
- Valkey is an open-source, community-driven project backed by the Linux Foundation, and Memorystore is described as a drop-in replacement for Redis OSS workloads.
- Memorystore for Valkey 9.0 delivers up to 3 times higher QPS at microsecond latency compared to Memorystore for Redis Cluster.
- Pipeline memory prefetching and zero-copy responses reduce CPU processing overhead.
- Hash Field Expirations and Atomic Slot Migration provide advanced cache control and zero-downtime cluster scaling.
- Vector Search supports semantic caching, dynamic LLM prompt memory, and real-time AI agents.
- Google Cloud offers an online migration workflow with continuous background replication and no downtime.
- Managed features include cluster provisioning, failovers, security patching, scaling, CMEK, granular ACLs, and AOF/RDB persistence.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:06 | Introducing AI-scale caching | Memorystore for Valkey is introduced as a highly available, scalable service for sub-millisecond data access. |
| 00:38 | Cache architecture and open-source compatibility | The video explains how Memorystore fits between applications and databases and offers compatibility with Redis OSS workloads. |
| 01:10 | Performance and engine capabilities | Valkey 9.0 is presented with performance gains, lower processing overhead, and features for cache control and cluster scaling. |
| 01:44 | Vector Search and migration | The video describes in-memory Vector Search for AI applications and an online migration process designed to avoid downtime. |
| 02:14 | Managed operations and reliability | Google Cloud manages deployment operations and provides enterprise features, four nines availability, and a call to start a migration. |
Quotes
Section titled “Quotes”Memorystore for Valkey delivers sub-millisecond data access, extreme scalability, and high availability — 00:06
Memorystore for Valkey 9.0 delivers up to 3 times higher QPS at microsecond latency — 01:10
Atomic Slot Migration for frictionless, fast, and zero-downtime cluster scaling. — 01:44
Google Cloud provides a fully managed, online migration workflow directly into Memorystore with no downtime. — 02:14
next-gen speed, total open-source freedom, and fully managed enterprise reliability. — 03:15
Concepts and entities
Section titled “Concepts and entities”Memorystore for Valkey, Valkey, Google Cloud, Redis OSS, Memorystore for Redis Cluster, Vector Search, Cloud SQL, AlloyDB, Spanner, Firestore, Linux Foundation
Full transcript (original language)
Section titled “Full transcript (original language)”00:06 Welcome to the era of AI-scale caching. When your application demands instant data responsiveness, Memorystore for Valkey delivers sub-millisecond data access, extreme scalability, and high availability as a fully-managed, 100% open-source in-memory database service. Memorystore is used by over 95% of the top 100 Google Cloud customers. These customers count on Memorystore to anchor their most essential applications when rapid performance is non-negotiable.
00:38 To achieve sub-millisecond responses, you can deploy Memorystore as an ‘in-memory cache’ positioned right between your application and your backend database— whether that is Cloud SQL, AlloyDB, Spanner, Firestore, or any other databases. Built on Valkey—the open-source, community-driven project backed by the Linux Foundation— Memorystore unlocks true licensing freedom and seamless compatibility as a drop-in replacement for your Redis OSS workloads. By keeping data right in memory, your apps can serve massive queries
01:10 This helps to drive engagement and effortless workflows with speed. I’m going to show you why modern engineering teams are moving to Memorystore for Valkey. Because it’s built on this open-source engine, teams gain total open-source freedom and are shielded from restrictive licensing shifts. This transition also delivers drastic price-performance gains and a fully managed cloud ecosystem. Let’s see what this means for your applications. Memorystore for Valkey 9.0 delivers up to 3 times higher QPS at microsecond latency compared to Memorystore for Redis Cluster.
01:44 Architectural optimizations like pipeline memory prefetching and zero-copy responses reduce CPU processing overhead, enabling your applications to stream massive data payloads straight out of memory without hitting execution bottlenecks. It also unlocks advanced engine capabilities like Hash Field Expirations for precise cache control, and Atomic Slot Migration for frictionless, fast, and zero-downtime cluster scaling. Are you building GenAI applications? Memorystore brings high-performance
02:14 Vector Search straight to your in-memory tier powering semantic caching, dynamic LLM prompt memory, and real-time AI agents. Migrating from self-managed Redis or Valkey is completely frictionless. Google Cloud provides a fully managed, online migration workflow directly into Memorystore with no downtime. Our automated tooling handles live replication continuously in the background, letting you flip production traffic over to Memorystore in seconds without missing a beat.
02:44 And once deployed, Google Cloud handles cluster provisioning, automated failovers, security patching, and vertical and horizontal scaling with zero downtime out of the box. You get enterprise-grade features including Customer-Managed Encryption Keys (CMEK), Granular ACLs, and AOF/RDB Persistence— all backed by four nines availability. That means saving hundreds of hours of manual cluster maintenance every year and redirecting your energy into launching innovative features faster.
03:15 Google Cloud Memorystore for Valkey gives you next-gen speed, total open-source freedom, and fully managed enterprise reliability. Ready to accelerate your apps and modernize your cache risk-free? Visit cloud.google.com/memorystore to start your migration or launch your first instance today.