| Azure SQL | Azure Cosmos | |
|---|---|---|
| Data Model | Relational tables, T-SQL, strong schema, ACID transactions. | Schemaless JSON documents (or MongoDB/Cassandra/Gremlin/Table models); multi-model, vector support. |
| Scale | “Scale up” (vCores/DTUs) with optional read-scale-out or geo-replicas. | “Scale out” automatically via physical partitions; virtually unlimited throughput & storage. |
| Consistency | Strict (snapshot, serializable, etc.). | Five tunable levels (Strong → Eventual). |
| Pricing unit | vCore / DTU / serverless per-second; long-running transactions encouraged. | Request Units (RU/s) for reads, writes & queries; optimize for small atomic operations. |
| When to pick | OLTP/OLAP apps that need joins, stored procs, mature relational tooling. | Globally distributed, high-throughput, low-latency micro-services, IoT, gaming, personalisation, etc. |
| Latency & SLA | Single-region HA SLA 99.99 %; write latency measured in ms – tens ms. | Multi-region (reads & writes) SLA 99.999 %; P99 <10 ms reads/writes in region. |
Sources: https://learn.microsoft.com/en-us/azure/azure-sql/database/?view=azuresql