Microsoft Events presents a practical guide to monitoring PostgresSQL and LLM workloads on Azure, featuring Datadog’s tooling and real-world troubleshooting tips for GenAI applications, led by Ryan MacLean.

Practical PostgreSQL and LLM Observability on Azure

Presented at Microsoft Ignite 2025 by Microsoft Events, with speaker Ryan MacLean, this intermediate session focuses on monitoring and optimizing GenAI and LLM workloads on Azure using Datadog.

Key Topics Covered

• Datadog's Recognition: As a Gartner Magic Quadrant leader, Datadog offers comprehensive Azure monitoring features.
• GenAI Scenario: Demonstrates real-life internal project issues, focusing on word similarity concepts, vector space representations, and AI tool integration on Azure.
• Vector Representation: Explains N-dimensional spaces and illustrates with examples like ‘Dog’ vs ‘Cat’ for word similarity within LLM contexts.
• Scaling Challenges: Addresses pitfalls such as blind spots in code across multiple teams, code testing under unpredictable real-world conditions, and database misconfigurations.
• Azure Deployment Issues: Examines downtime causes, ways to investigate SQL query trace and performance problems, and highlights actionable methods for debugging in Azure environments.

Azure Monitoring with Datadog

• Step-by-step guidance for integrating Datadog with Azure.
• How to get started with PostgreSQL on Azure, track relevant performance metrics for GenAI workloads, and leverage real-time dashboards.
• Creation of alerts providing meaningful insights into operational health, query performance, and database misconfigurations.

Troubleshooting and Best Practices

• Identifying deployment errors leading to downtime.
• Approaches for monitoring distributed systems and AI applications at scale.
• Practical tips for debugging blind spots in code and ensuring reliability in dynamic environments.

Who Should Watch?

• Developers and DevOps engineers deploying GenAI, LLMs, PostgreSQL, and other critical workloads on Azure.
• Technical teams interested in unified observability and actionable analytics for cloud databases and AI tooling.

Further Learning

Explore additional sessions and resources on Microsoft Ignite at ignite.microsoft.com.