From Rising RU Costs to Stable Performance | Anurag Dutt | Azure Cosmos DB Conf 2026

As Cosmos DB workloads scale, teams often experience rising RU costs, unpredictable latency, and uneven throughput caused by early design decisions that no longer meet production requirements. Anurag Dutt (Multi-Cloud Engineer) walks through a realistic, high-volume data ingestion and querying workload that experienced rapid cost growth and performance degradation over time. Using Azure platform metrics, diagnostics, and query insights, he identifies hot partitions, cross-partition queries, over-indexing, and inefficient throughput provisioning — then shows the targeted improvements that restored stability and cut costs. You'll leave with a practical, repeatable framework for diagnosing and resolving common Cosmos DB performance and cost challenges. 👤 Connect with Anurag Dutt 📝 Anurag is a Cloud Engineer with expertise in multi-cloud architectures, automation, and infrastructure management across AWS, Azure, and GCP. Passionate about cloud-native technologies and open-source, he helps teams optimize cloud workloads for scalability, security, and cost efficiency. With hands-on experience in IaC, Kubernetes, and DevOps, Anurag simplifies complex cloud concepts into practical, real-world strategies for developers and engineers. 💼 https://www.linkedin.com/in/anuragdutt01/ 𝕏 https://x.com/duttanurag20 🎤 https://developer.azurecosmosdb.com/conf/#speaker/anurag-dutt --- Azure Cosmos DB Conf 2026 is a free virtual developer event focused on building modern, scalable applications with Azure Cosmos DB. Sessions feature Microsoft engineers, product teams, and members of the global developer community sharing real-world solutions and best practices. Originally streamed on April 28, 2026 | 9:00 AM – 2:00 PM PT 🔗 Useful Links • Explore all sessions on-demand — https://aka.ms/CosmosConf26Playlist • Take the Cosmos Conf Challenge. First 500 to complete gets a voucher for 100% free certificate exam for DP-420 — https://aka.ms/CosmosDBConfChallenge • Share your feedback