Skip to content

Azure Data Services

Scope

Azure managed data platform services. Covers Azure SQL (vCore/DTU/Hyperscale), Cosmos DB (consistency levels, multi-region), Azure Cache for Redis tiers, geo-replication, Private Endpoints, and managed identity authentication.

Checklist

  • [Critical] Is Azure SQL Database chosen for relational workloads, with the appropriate purchasing model? (vCore for predictable performance, DTU for simpler workloads, Hyperscale for large databases)
  • [Critical] Is Azure SQL configured with zone-redundant high availability (available in Premium, Business Critical, and Hyperscale tiers)?
  • [Critical] Is auto-failover groups configured for Azure SQL when cross-region disaster recovery is required?
  • [Critical] Is Cosmos DB selected for globally distributed, multi-model workloads, with the appropriate consistency level chosen? (strong, bounded staleness, session, consistent prefix, eventual)
  • [Recommended] Are Cosmos DB Request Units (RUs) right-sized, with autoscale configured for variable workloads and provisioned throughput for predictable workloads?
  • [Recommended] Is Cosmos DB multi-region write enabled only when the application can handle conflict resolution, otherwise using single-write with read replicas?
  • [Recommended] Is Azure Cache for Redis deployed in private subnets with the appropriate tier? (Standard for basic caching, Premium for persistence/clustering, Enterprise for RediSearch/RedisBloom)
  • [Recommended] Is Azure Cache for Redis zone-redundant (Premium and Enterprise tiers) for production workloads?
  • [Critical] Are geo-replication configurations tested for failover and failback procedures with documented RTO/RPO?
  • [Critical] Is Transparent Data Encryption (TDE) enabled with customer-managed keys for Azure SQL and Cosmos DB?
  • [Critical] Are Private Endpoints configured for all data services to prevent public network access?
  • [Critical] Are database connections using managed identity authentication where supported instead of connection strings with passwords?
  • [Optional] Is Azure SQL Intelligent Performance (automatic tuning, Query Performance Insight) enabled for self-tuning and diagnostics?
  • [Recommended] Are long-term backup retention (LTR) policies configured for compliance, with backups stored in a separate region?

Why This Matters

Azure data services have unique pricing models (DTU vs vCore, Request Units) that significantly affect cost. Cosmos DB consistency level selection directly impacts performance, availability, and correctness. Azure SQL tiers determine available features like zone redundancy and read scale-out. Misconfigured geo-replication creates false confidence in disaster recovery capability.

Common Decisions (ADR Triggers)

  • Azure SQL vs Cosmos DB -- relational with transactions vs globally distributed multi-model
  • Azure SQL purchasing model -- vCore (resource control) vs DTU (bundled) vs Hyperscale (100TB+, rapid scale)
  • Cosmos DB API selection -- NoSQL (native) vs PostgreSQL vs MongoDB vs Cassandra vs Gremlin vs Table
  • Cosmos DB consistency level -- strong (highest latency) vs session (most popular) vs eventual (lowest latency)
  • Caching tier -- Azure Cache for Redis Standard vs Premium (clustering, persistence) vs Enterprise (Redis modules)
  • Azure SQL Managed Instance vs Azure SQL Database -- near-100% SQL Server compatibility vs fully managed PaaS
  • Serverless compute -- Azure SQL Serverless with auto-pause for intermittent workloads vs provisioned for always-on

Reference Architectures

See Also

Looking for analytics, data lake, or data warehouse guidance? This file covers Azure operational databases only. For the analytics layer see providers/azure/fabric.md (Microsoft Fabric, OneLake, lakehouse and warehouse, Direct Lake, capacity economics), providers/azure/analytics.md (Synapse SQL and Spark pools, Data Factory, Event Hubs, Stream Analytics, Synapse Link), and providers/azure/data-explorer.md (telemetry and time-series).

  • providers/azure/fabric.md -- Microsoft Fabric and OneLake: the analytics and lakehouse platform, and the target for Cosmos DB and Azure SQL mirroring
  • providers/azure/analytics.md -- Azure Synapse Analytics, Data Factory, Event Hubs, Stream Analytics, and Synapse Link for Cosmos DB HTAP
  • providers/azure/data-explorer.md -- Azure Data Explorer for telemetry, log, and time-series analytics
  • general/data.md -- General data architecture patterns and database selection criteria
  • general/data-governance.md -- Catalog, lineage, classification, and access governance across data services
  • providers/azure/database.md -- Detailed Azure database service coverage including PostgreSQL and MySQL
  • providers/azure/cosmos-db.md -- Cosmos DB modelling, partitioning, and consistency in depth
  • providers/azure/networking.md -- Private Endpoints for securing database network access
  • providers/azure/disaster-recovery.md -- Geo-replication and failover strategies for Azure data services