Azure Managed Redis & Azure Cosmos DB with cache‑aside: a practical guide
Co-authored by James Codella - Principal Product Manager, Azure Cosmos DB, Microsoft; Andrew Liu - Principal Group Product Manager, Azure Cosmos DB, Microsoft; Philip Laussermair – Azure Managed Redis Solution Architect, Redis Inc. Using Azure Managed Redis alongside Azure Cosmos DB is a powerful way to reduce operational costs in read-heavy applications. While Azure Cosmos DB delivers low-latency point reads, an SLA-backed 10ms at the 99th percentile of requests, each read consumes Request Units (RUs), which directly impact your billing. For workloads with frequent access to the same data, caching those reads in Azure Managed Redis can dramatically reduce RU consumption and smooth out cost spikes. This cache-aside pattern allows applications to serve hot data from memory while preserving Azure Cosmos DB as the system of record. By shifting repeated reads to Azure Managed Redis, developers can optimize for cost efficiency without sacrificing consistency, durability, or global availability. What each service does: Azure Managed Redis is Microsoft’s first‑party, fully managed service built on Redis Enterprise. Azure Cosmos DB offers higher durable multi-region writes, high availability SLAs, and in-region single-digit millisecond latency for point operations. Co-locating app compute, Azure Managed Redis, and Azure Cosmos DB in the same region minimizes round-trips; adding Azure Managed Redis on top of Azure Cosmos DB reduces RU consumption from repeat reads and smooths tail latency spikes during peak load. Both Azure Managed Redis and Azure Cosmos DB offer great support for queries (including vector search) over JSON data to work well together fast efficient AI apps. This pairing doesn’t require a rewrite. You can adopt a cache‑aside strategy in your data‑access layer: look in Redis first; on a miss, read from Azure Cosmos DB and populate Redis with a TTL; on writes, update Azure Cosmos DB and invalidate or refresh the corresponding cache key. Use Azure Cosmos DB ETags in cache keys to make invalidation deterministic, and use the Azure Cosmos DB Change Feed to trigger precise cache refreshes when data changes. The Cache‑Aside (Lazy Loading) Pattern Read path: GET key from Azure Managed Redis. If found, return. If not, issue a point read to Azure Cosmos DB, then SET/JSON.SET the value in Azure Managed Redis with a TTL and return the payload to the caller. Write path: Persist to Azure Cosmos DB as the source of truth. Invalidate or refresh the related Redis key (for example, delete product:{id}:v{etag} or write the new version). If you subscribe to the Change Feed, an Azure Function can perform this invalidation asynchronously to keep caches hot under write bursts. Code Example (.NET): using System; using System.Collections.Generic; using System.Threading.Tasks; using Microsoft.Azure.Documents; using Microsoft.Extensions.Logging; using StackExchange.Redis; using Microsoft.Azure.StackExchangeRedis; using Azure.Identity; public static class CosmosDbChangeFeedFunction { private static RedisConnection _redisConnection; static CosmosDbChangeFeedFunction() { // Initialize Redis connection using Entra ID (Azure AD) authentication var redisHostName = Environment.GetEnvironmentVariable("RedisHostName"); // e.g., mycache.redis.cache.windows.net var credential = new DefaultAzureCredential(); var configurationOptions = ConfigurationOptions.Parse($"{redisHostName}:10000"); _redisConnection = RedisConnection.ConnectAsync(configurationOptions.ConfigureForAzureWithTokenCredentialAsync(credential)).GetAwaiter().GetResult(); } [FunctionName("CosmosDbChangeFeedFunction")] public static async Task Run( [CosmosDBTrigger( databaseName: "my-database", containerName: "my-container", ConnectionStringSetting = "CosmosDBConnection", LeaseContainerName = "leases")] IReadOnlyList<Document> input, ILogger log) { var cache = _redisConnection.GetDatabase(); if (input != null && input.Count > 0) { foreach (var doc in input) { string id = doc.GetPropertyValue<string>("id"); string etag = doc.GetPropertyValue<string>("_etag"); string cacheKey = $"item:{id}:v{etag}"; string json = doc.ToString(); await cache.StringSetAsync(cacheKey, json, TimeSpan.FromMinutes(10)); log.LogInformation($"🔄 Refreshed cache for key: {cacheKey}"); } } } } Why it works: The database handles correctness and global replication; the cache handles locality and frequency. You reduce repeated reads (and RU costs) and lower p99 by serving hot keys from memory close to the compute. TTLs give you explicit control over staleness; negative caching and stale‑while‑revalidate are easy extensions when appropriate. Design Choices That Matter TTLs: Choose TTLs that reflect business tolerance for staleness. Use jitter (±N%) to avoid thundering herds when many keys expire simultaneously. Keying and versioning: Include an ETag or version in the key, e.g., product:{id}:v{etag}. When the record changes, the ETag changes, naturally busting the old key. Cache stampede control: For hot keys that miss, use a short single‑flight lock so that one request refreshes the value while others wait briefly or serve stale data. Serialization: For portability, utilize compact JSON (RedisJSON if you want field‑level reads/writes). Keep values small to preserve cache efficiency. Failure semantics: Treat Azure Managed Redis as an optimization. If the cache is unavailable, the app should continue by reading from Azure Cosmos DB. Favor idempotent writes and retry‑safe operations. Why Azure Managed Redis & Azure Cosmos DB Work Well in Practice Local speed, global reach: Azure Cosmos DB targets a single-digit millisecond p99 for in-region point operations. Placing Azure Managed Redis in the same region enables sub-millisecond memory reads for repeated access patterns, providing optimal performance for high-frequency data access. The result is a shorter, more predictable critical path. Active‑active at both layers: Azure Cosmos DB supports multi‑region writes, so each region can accept traffic locally. Azure Managed Redis supports active geo‑replication across up to five instances, using conflict-free replicated data types (CRDT) to converge cache state. That yields region‑local cache hits with eventual consistency at the cache tier and strong guarantees at the database tier. Reference Architecture Regional deployment Ingress: Clients reach the app via Azure Front Door (or similar) and land on Azure App Service in a VNet. Read path: The app queries Azure Managed Redis first. On a miss, it performs a point read/write against Azure Cosmos DB (API for NoSQL) and updates Redis with a TTL. Write path: Writes go to Azure Cosmos DB. The Change Feed triggers an Azure Function that invalidates or refreshes related Redis keys. Observability: Use Azure Monitor for logs and metrics. Monitor cache hit ratio, gets/sets, evictions, and geo‑replication health on the Azure Managed Redis side; RU/s, throttles, and p99 latency on the Azure Cosmos DB side. Operations and SRE Considerations Co‑location: Keep compute, Azure Managed Redis, and the Azure Cosmos DB write/read region together to avoid unnecessary RTTs. Capacity planning: Size Azure Managed Redis memory for working‑set coverage and headroom for failover. Validate eviction policies (volatile‑TTL vs all‑keys) against workload behavior. Back‑pressure: Watch RU throttling on Azure Cosmos DB and evictions on Azure Managed Redis. High evictions or a low hit ratio indicate a working-set mismatch or TTLs that are too short. Testing: Load‑test with realistic key distributions and measure p50/p95/p99 on both cache hits and misses. Chaos‑test cache outages to verify graceful degradation. Security: Use managed identities for data plane access where supported; apply App Service access restrictions and VNet integration as appropriate; audit with Azure Monitor logs. Putting It All Together Adopt cache‑aside in one region, measure hit ratio and RU savings, then add Change Feed–based invalidation to keep hot keys fresh under write load. When you need global scale, enable Azure Cosmos DB multi-region writes and Azure Managed Redis active geo-replication so that every region serves users locally. You end up with fast read paths, clear consistency boundaries, and a deployment model that scales without surprises. Next steps Review Azure Managed Redis documentation - https://learn.microsoft.com/azure/redis/ , Learn more about Cache Aside pattern in the Azure Architecture Center - https://learn.microsoft.com/azure/architecture/patterns/cache-aside Start with a sample app - https://github.com/AzureManagedRedis/What’s new in Azure Managed Redis: Ignite 2025 feature announcements
Azure Managed Redis continues to power the world’s most demanding, low-latency workloads—from caching and session management to powering the next generation of AI agents. At Microsoft Ignite 2025, we’re excited to announce several new capabilities designed to make Azure Managed Redis even more scalable, manageable, and AI-ready. Bigger, faster, stronger: new enterprise-scale SKUs (generally available) We are expanding our capacity portfolio with the general availability of Memory Optimized 150 and 250, Balanced 150 and 250, Compute Optimized 150 and 250 SKUs/, bringing higher throughput, lower latency, and greater memory capacity for your most demanding workloads. Whether you’re running global gaming platforms, AI-powered personalization engines, or enterprise-scale caching tiers, these new SKUs offer the performance headroom to scale with confidence. Redis as a knowledge base for AI Agents Azure Managed Redis is now available as part of the Azure AI Foundry MCP tools catalog, allowing customers to use Redis as a knowledge store or memory store for AI agents. This integration makes it simple to connect Redis to Foundry-based agents, enabling semantic searching, long term and short term memory, and faster reasoning for multi-agent applications—all running on trusted Azure infrastructure. Scheduled Maintenance (public preview) You can now configure maintenance windows for their Azure Managed Redis instances, giving you greater control and predictability for planned service updates. This capability helps align maintenance with your own operational schedules—minimizing disruption and providing flexibility for mission-critical applications. Terraform Provider for Azure Managed Redis We’re making infrastructure automation even easier with a dedicated Terraform provider for Azure Managed Redis. This new provider enables you to declaratively create, manage, and configure AMR resources through code, improving consistency and streamlining CI/CD pipelines across environments. Reserved Instances: now in 30+ Regions Azure Managed Redis now supports Reserved Instances in over 30 regions, with more coming soon. Reserved pricing provides predictable costs and savings for long-term workloads.. Go to Azure Portal | Reservations | Add and search for ‘Azure Cache for Redis’. Azure Managed Redis SKUs like Balanced, Computer Optimized, Memory Optimized, and Flash Optimized, would show up as sub SKUs in the category. Reserved Instance is available for 35% discount with 1 year purchase and 55% discount with 3 year purchase. Learn More & Get Started Azure Managed Redis is redefining what’s possible for caching, data persistence, and agentic AI workloads. Explore the latest demos, architecture examples, and tutorials from Ignite: Learn more about Azure Managed Redis Try Azure Managed Redis samples on GitHub Watch the Azure Managed Redis session at Ignite on-demand (BRK129) Explore Redis as a memory store for Microsoft Agent Framework Introducing the PublicNetworkAccess property to Azure Managed Redis | Microsoft Community Hub Ready to build Internet-scale AI apps with Azure Managed Redis? Start today at aka.ms/hol-amr.Azure Managed Redis at Ignite 2025: pre-day, session, and booth
Microsoft Ignite 2025 is almost here! Many practitioners are surprised by the powerful new capabilities in Azure Managed Redis—and now is your chance to see them in action. Whether you are modernizing applications, accelerating AI workloads, or building next-generation agent architectures, Azure Managed Redis is your key to speed and scale. Don’t miss the chance to connect with experts from Microsoft and Redis at our pre-day workshop and general session at Ignite and learn how to: Unlock high-performance caching for demanding workloads Build a powerful memory layer for agentic applications Leverage vector storage for Retrieval-Augmented Generation (RAG) Optimize LLM costs with semantic caching All in one fully-managed service—Azure Managed Redis. Connect with Azure Managed Redis team at Ignite 2025 1. Ignite pre-day workshop: Build Internet-Scale AI Apps with Azure Managed Redis — Caching to Agents (in-person only) When: Ignite pre-day on Monday, November 17, 2025, 1pm-5pm PT Where: Moscone Center, San Francisco Registration: Add this optional in-person workshop in the Ignite registration → Building AI applications and agents at internet scale requires more than speed — it demands unified memory, context, and scalability. You’ll see live demos and learn how to build and scale intelligent applications using Azure Managed Redis for caching and modern AI workloads, architect your applications for performance, reliability, and scale with geo-replication, and migrate to Azure Managed Redis. Seats are limited, please sign up today. 2. Breakout Session: Smarter AI Agents with Azure Managed Redis - BRK129 (in-person and online) View session details on the Ignite website and save to your event favorites Azure Managed Redis with Azure AI Foundry and Microsoft Agent Framework let developers build adaptive, context-aware AI systems. Redis handles real-time collaboration, persistent learning, and semantic routing, while the Agent Framework supports advanced reasoning and planning. Integrated short- and long-term memory lets agents access relevant data directly, simplifying development and operations. Azure Managed Redis supports MCP for control and data plane tasks, enabling easy management and scaling of workloads and knowledge stores. Join us to discover how to build scalable, multi-agent systems backed by the performance, reliability, and unified memory of Redis on Azure. 3. Visit the Azure Managed Redis booth in the Expo Hall Have questions? Looking to talk architecture, migration, and supercharging your AI apps? Visit us in the Expert Meetup Zone to connect with the Microsoft and Redis product teams, engineers, and architects behind Azure Managed Redis. Prepare for Ignite Learn more about Microsoft Ignite Explore the Azure Managed Redis documentation. Try the hands-on workshop for Azure Managed Redis.Redis named Stack Overflow’s top data storage tool for AI Agents
Redis recognized by developers on Stack Overflow as the #1 data storage tool for AI agent workloads and #5 database in 2025. Microsoft is the only cloud provider to offer Redis Enterprise as a fully native, managed service with Azure Managed Redis.
Orchestrate multi-LLM workflows with Azure Managed Redis
Authors: Roberto Perez, George von Bülow & Roy de Milde Key challenge for building effective LLMs In the age of generative AI, large language models (LLMs) are reshaping how we build applications — from chatbots to intelligent agents and beyond. But as these systems become more dynamic and multi-modal, one key challenge stands out: how do we route requests efficiently to the right model, prompt, or action at the right time? Traditional architectures struggle with the speed and precision required to orchestrate LLM calls in real-time, especially at scale. This is where Azure Managed Redis steps in — acting as a fast, in-memory data layer to power smart, context-aware routing for LLMs. In this blog, we explore how Redis and Azure are enabling developers to build AI systems that respond faster, think smarter, and scale effortlessly. Across industries, customers are hitting real limitations. AI workloads often need to track context across multiple interactions, store intermediate decisions, and switch between different prompts or models based on user intent — all while staying responsive. But stitching this logic together using traditional databases or microservice queues introduces latency, complexity, and cost. Teams face challenges like keeping routing logic fast and adaptive, storing transient LLM state without bloating backend services, and coordinating agent-like behaviors across multiple components. These are exactly the pain points AMR was built to address — giving developers a low-latency, highly available foundation for real-time AI orchestration and more. How to use Azure Managed Redis as a Semantic Router Semantic routing uses AI to route user queries to the right service, model or endpoint, based on their intent and context. Unlike rule-based systems, it leverages Generative AI to understand the meaning behind requests, enabling more accurate and efficient decisions. Importantly, the semantic router itself does not forward the query—it only selects the appropriate route. Your application is responsible for taking that routing decision and sending the query to the correct agent, model, or human. The users sends a query, which is passed to the system for processing The query is analyzed by an embedding model to understand its semantic intent and context The semantic router evaluates the user’s intent and context to choose the optimal route: A specific model for further processing An agent to handle the query A default response if applicable Escalation to a human for manual handling, if needed Valid queries go through the RAG pipeline to generate a response The final response is sent back to the user Code examples + Architecture Example: Jupyter Notebook with Semantic Router Let’s look at a Jupyter Notebook example that implements a simple Semantic Router with Azure Managed Redis and the Redis Vector Library. First, we install the required Python packages and define a connection to an AMR instance: pip install -q "redisvl>=0.6.0" sentence-transformers dotenv Define the Azure Managed Redis Connection. import os import warnings warnings.filterwarnings("ignore") from dotenv import load_dotenv load_dotenv() REDIS_HOST = os.getenv("REDIS_HOST") # ex: "gvb-sm.uksouth.redis.azure.net" REDIS_PORT = os.getenv("REDIS_PORT") # for AMR this is always 10000 REDIS_PASSWORD = os.getenv("REDIS_PASSWORD") # ex: "giMzOzIP4YmjNBGCfmqpgA7e749d6GyIHAzCaF5XXXXX" # If SSL is enabled on the endpoint, use rediss:// as the URL prefix REDIS_URL = f"redis://:{REDIS_PASSWORD}@{REDIS_HOST}:{REDIS_PORT}" Next, we create our first Semantic Router with an allow/block list: from redisvl.extensions.router import Route, SemanticRouter from redisvl.utils.vectorize import HFTextVectorizer vectorizer = HFTextVectorizer() # Semantic router blocked_references = [ "things about aliens", "corporate questions about agile", "anything about the S&P 500", ] blocked_route = Route(name="block_list", references=blocked_references) block_router = SemanticRouter( name="bouncer", vectorizer=vectorizer, routes=[blocked_route], redis_url=REDIS_URL, overwrite=False, ) To prevent users from asking certain categories of questions, we can define example references in a list of blocked routes using the Redis Vector Library function SemanticRouter(). While it is also possible to implement blocking at the LLM level through prompt engineering (e.g., instructing the model to refuse answering certain queries), this approach still requires an LLM call, adding unnecessary cost and latency. By handling blocking earlier with semantic routing in Azure Managed Redis, unwanted queries can be intercepted before ever reaching the model, saving LLM tokens, reducing expenses, and improving overall efficiency. Let’s try it out: user_query = "Why is agile so important?" route_match = block_router(user_query) route_match The router first vectorizes the user query using the specified Hugging Face text vectorizer. It finds a semantic similarity with route reference “corporate question sabout agile” and returns the matching route ‘block_list`. Note the returned distance value – this indicates the degree of semantic similarity between the user query and the blocked reference. You can fine-tune the Semantic Router by specifying a minimum threshold value that must be reached to count as a match. For full details and more complex examples, you can explore the Jupyter Notebooks in this GitHub repository. How do customers benefit? For customers, this technology delivers clear and immediate value. By using Azure Managed Redis as the high-performance backbone for semantic routing and agent coordination, organizations can significantly reduce latency, simplify infrastructure, and accelerate time-to-value for AI-driven experiences. Instead of building custom logic spread across multiple services, teams get a centralized, scalable, and fully managed in-memory layer that handles vector search, routing logic, and real-time state management — all with enterprise-grade SLAs, security, and Azure-native integration. The result? Smarter and faster LLM interactions, reduced operational complexity, and the flexibility to scale AI use cases from prototypes to production without re-architecting. Whether you're building an intelligent chatbot, orchestrating multi-agent workflows, or powering internal copilots, this Redis-backed technology gives you the agility to adapt in real time. You can dynamically route based on user intent, past interactions, or even business rules — all while maintaining low-latency responses that users expect from modern AI applications. And because it’s fully managed on Azure, teams can focus on innovation rather than infrastructure, with built-in support for high availability, monitoring, and enterprise governance. It’s a future-proof foundation for AI systems that need to be not just powerful, but precise. Try Azure Managed Redis today If you want to explore how to route large language models efficiently, Azure Managed Redis provides a reliable and low-latency solution. You can learn more about the service on the Azure Managed Redis page and find detailed documentation in the Azure Redis overview. For hands-on experience, check out the routing optimization notebook and other examples in the Redis AI resources repository and GitHub - loriotpiroloriol/amr-semantic-router. Give it a try to see how it fits your LLM routing needs.Building faster AI agents with Azure Managed Redis and .NET Aspire
AI is evolving fast—and so are the tools to build intelligent, responsive applications. In our recent Microsoft Reactor session, Catherine Wang (Principal Product Manager at Microsoft) and Roberto Perez (Microsoft MVP and Senior Global Solutions Architect at Redis) shared how Azure Managed Redis helps you create Retrieval-Augmented Generation (RAG) AI agents with exceptional speed and consistency. Why RAG agents? RAG applications combine the power of large language models (LLMs) with your own data to answer questions accurately. For example, a customer support chatbot can deliver precise, pre-approved answers instead of inventing them on the fly. This ensures consistency, reduces risk, and improves customer experience. Where Azure Managed Redis fits with agentic scenarios In this project, Azure Managed Redis is used as a high-performance, in-memory vector database to support Agentic Retrieval-Augmented Generation (RAG), enabling fast similarity searches over embeddings to retrieve and ground the LLM with the most relevant known answers. Beyond this, Azure Managed Redis is a versatile platform that supports a range of AI-native use cases, including: Semantic Cache – Cache and reuse previous LLM responses based on semantic similarity to reduce latency and improve reliability. LLM Memory – Persist recent interactions and context to maintain coherent, multi-turn conversations. Agentic Memory – Store long-term agent knowledge, actions, and plans to enable more intelligent and autonomous behavior over time. Feature Store – Serve real-time features to machine learning models during inference for personalization and decision-making. These capabilities make Azure Managed Redis a foundational building block for building fast, stateful, and intelligent AI applications. Demo highlights In the session, the team demonstrates how to: Deploy a RAG AI agent using .NET Aspire and Azure Container Apps. Secure your Redis instance with Azure Entra ID, removing the need for connection strings. Use Semantic Kernel to orchestrate agents and retrieve knowledge base content via vector search. Monitor and debug microservices with built-in observability tools. Finally, we walk through code examples in C# and Python, demonstrating how you can integrate Redis search, vector similarity, and prompt orchestration into your own apps. Get Started Ready to explore? ✅ Watch the full session replay: Building a RAG AI Agent Using Azure Redis ✅ Try the sample code: Azure Managed Redis RAG AI SampleGet started with Azure Managed Redis today: a step-by-step guide to deployment
At Microsoft Build 2025, we announced the general availability of Azure Managed Redis, a fully-managed, first-party service built in partnership with Redis. Ready for production workloads globally, Azure Managed Redis marks a major milestone for developers looking to build high-performance, real-time applications with the speed and reliability of Redis, fully managed on Azure. Call to action: get started with Azure Managed Redis in the Azure Portal. Key updates: Up to 15x performance improvements over Azure Cache for Redis 99.999% availability with multi-region Active‑Active replication Support for Redis 7.4 (with Redis 8 coming soon) New modules including RedisJSON, vector search, bloom filters, and time-series Flexible SKUs that let you scale memory and compute independently Navigate the new Azure Managed Redis in the Azure Portal Azure Managed Redis also comes with an updated Azure Portal experience which simplifies how you create, configure, and manage your Redis instances. Whether experimenting or deploying to production, the portal gives you full control with a few clicks. Step-by-step guide to deploying in the Azure Portal Want to see Azure Managed Redis in action? This quick walkthrough video shows how to set up Azure Managed Redis inside the Azure Portal: 👉 Watch on YouTube In this tutorial, you’ll learn how to: How to configure your Active-Active instance for high availability and low latency Setting up geo-replication across regions for 99.999% availability SLA Key tips and best practices to get started quickly No code required — just the Azure Portal and a few minutes of your time! Azure Managed Redis is perfect for cloud architects, developers, and IT pros looking to build resilient, globally available Redis-backed applications on Azure. Whether you're building AI-powered applications, speeding up your web services, or just getting started with Redis, now’s the time to explore what Azure Managed Redis can do. To learn more, head to our product page for more information or contact your Microsoft sales representative. To get started, provision Azure Managed Redis in the Azure Portal today. Resources Azure Managed Redis product page Azure Managed Redis pricing page Create an Azure Managed Redis instance Watch the Microsoft Build 2025 session on AMR Explore the documentation
