Azure Update 27th March 2026

Key Concepts

• Azure Kubernetes Service (AKS): Managed Kubernetes service with new networking, observability, and deployment features.
• Vector Search (DiskANN): Microsoft Research-developed algorithm for high-dimensional vector search in SQL databases.
• Microsoft Fabric Mirroring: Near real-time data replication from operational databases (MySQL, Cosmos DB) to OneLake.
• Priority Processing: A pay-as-you-go model for AI inferencing that provides lower latency and higher throughput without pre-provisioned throughput units (PTUs).
• Entra Tenant Governance: Tools for identifying shadow tenants and enforcing secure configuration at the point of creation.

---

1. Compute and AKS Updates

• Application Network (Preview): A new abstraction layer for Kubernetes traffic control. It enables service-to-service communication management and observability without sidecars, utilizing SPIFFE (Secure Production Identity Framework for Everyone) for identity-based network control.
• Network Logs (GA): Provides metadata (IPs, ports, namespaces, flow direction) at layers 3, 4, and 7. Supports filtering to reduce noise and storage costs, with an on-demand mode using Hubble.
• Managed GPU Metrics (Preview): Integration with Managed Prometheus and Grafana to track Nvidia GPU performance (utilization, memory, temperature, power, and bandwidth).
• Fleet Manager Cross-Cluster Networking (Preview): A managed Cilium Cluster Mesh that allows services in one cluster to be accessed by another as if they were local, including global observability.
• Blue-Green Agent Pool Upgrades (Preview): A safe deployment methodology where a parallel node pool is created with new configurations. Traffic is shifted to the new pool; if successful, the old pool is deleted. Note: Requires double the quota/resources during the upgrade window.
• Azure Container Storage: Now supports Elastic SAN (GA), providing durable, flexible storage pools alongside existing local NVMe options.

2. Database and Data Services

• SQL Database Automatic Index Compaction (Preview): A background process for SQL DB, SQL MI, and SQL in Fabric that reduces storage footprint and improves performance by eliminating the need for manual index maintenance jobs.
• SQL Managed Instance Change Event Streaming (Preview): Streams row-level changes (INSERT/UPDATE/DELETE) to Event Hubs in near real-time, enabling event-driven architectures.
• SQL Hyperscale New SKUs (Preview): Introduction of 160 and 192 vCore options for premium hardware, targeting large-scale OLTP and HTAP workloads.
• DiskANN Improvements: Enhancements to vector search in SQL databases, including non-read-only tables post-index creation and performance-optimized filtering during searches.
• Azure Monitor OLTP Ingestion (Preview): Native support for OpenTelemetry protocols, allowing direct ingestion of metrics, logs, and traces into Azure Monitor workspaces using Entra authentication.
• Fabric Mirroring: MySQL Flexible Server and Cosmos DB (with private endpoints) can now mirror data to Fabric OneLake without complex data pipelines.

3. AI and Miscellaneous

• Priority Processing (GA): Offers high-speed, low-latency inferencing on a pay-as-you-go basis. It serves as an alternative or supplement to Provisioned Throughput Units (PTUs) for time-critical AI applications, albeit at a price premium.
• Entra ID External MFA (GA): Supports external MFA solutions via OpenID Connect within Entra ID authentication, replacing deprecated custom controls.
• Entra Tenant Governance (GA):
  • Shadow Tenant Detection: Identifies unauthorized or unknown tenants based on identity patterns and billing.
  • Secure Tenant Creation: Ensures new tenants adhere to organizational security policies from inception.

---

4. Synthesis and Conclusion

The updates from March 27th reflect a strong focus on operational efficiency and observability. In the Kubernetes space, the shift toward "meshless" networking and AI-driven troubleshooting (Network AI Agent) highlights a move to reduce management overhead. For data professionals, the integration of near real-time mirroring into Microsoft Fabric and the advancement of vector search capabilities (DiskANN) position Azure as a robust platform for generative AI and real-time analytics. Finally, the introduction of Priority Processing for AI inferencing provides developers with more flexible, performance-oriented cost models for scaling intelligent applications.