How Veeam Delivers Planet-Scale Semantic Search with Azure Cosmos DB
Microsoft Events shares insights from the Azure Cosmos DB engineering team and Veeam on using Cosmos DB for large-scale semantic search, detailing architecture, demos, and optimization strategies.
How Veeam Delivers Planet-Scale Semantic Search with Azure Cosmos DB
Presented by the Azure Cosmos DB engineering team and Veeam at Microsoft Ignite 2025
Overview
This session explores how Azure Cosmos DB enables enterprise-grade, scalable semantic search, focusing on new features for developers, AI integration, and high-performance search operations.
Key Topics Covered
Introduction to Microsoft DiskANN Vector Index for Semantic Search:
- Explains the core technology powering contextually relevant vector search within Azure Cosmos DB.
Hyper Scale Vector Search with Garnet Cache:
- Details about leveraging Garnet Cache to support high-throughput, low-latency vector searches at scale.
Security, Scalability, Reliability, and Compliance:
- Discussion of strategies to ensure regional compliance and secure data handling across global deployments.
System Architecture:
- How Azure Event Hubs, Azure Container Apps, Databricks, and Cosmos DB work together in an event-driven pipeline for data refinement, enrichment, and indexing.
- Breakdown of multi-tenant I/O patterns and shard key strategies to optimize search performance.
Demonstrations
- Veeam Data Cloud Admin Console & Project Atlas:
- Showcases a practical implementation of semantic search, including:
- Real-world demo of searching contextually relevant emails.
- Hybrid search that blends vector and text-based ranking approaches.
- Showcases a practical implementation of semantic search, including:
Development and Optimization Insights
- Multi-tenant data handling
- Efficient vector search architecture
- Shard keys for performance
- Data enrichment and indexing best practices
Why Use Azure Cosmos DB for Semantic Search?
- Unified solution for reliable storage, vector search, and full-text indexing
- Designed for global scale, enterprise reliability, and rapid search
- Built-in integrations with event-driven Microsoft Azure services
Additional Resources
Speakers
- James Codella
- Johan Huttenga
- James Nguyen
- Zachary Rossman
For more detailed architectural diagrams, implementation code, or follow-up Q&A, refer to the session recording and related links from Microsoft Ignite.