Basic Tier
- Max size: 2 GB
- Self Service Restore window: 7 days
- Geo replication: Active
- Performance level: Basic
Standard Tier
- Max size: 250 GB
- Self Service Restore window: 35 days
- Geo replication: Active
- Performance level: S0 S1 S2 S3
Premium Tier
- Max size: 500 GB (P1 - P6) or 1024 GB (P11 and P15 only)
- Self Service Restore window: 35 days
- Geo replication: max 4 readable secondaries
- Performance level: P1 P2 P4 P6 P11 P15
All Tiers
- SLA: 99.99%
- Max database servers: 6 per Azure Account
soft limit - Max databases per server: 150 (including master)
soft limit - Max number of database backup schedules: 10 per subscription
Performance Levels
| Service Tier/Performance Level | DTU [1] | Max Worker Threads | Max Sessions | Benchmark Transaction Rate [2] | Predictability [3] |
|---|---|---|---|---|---|
| Basic | 5 | 20 | 100 | 3,467 transactions/hour | Good |
| Standard/S0 | 10 | 50 | 200 | 283 transactions/minute | Better |
| Standard/S1 | 20 | 50 | 200 | 283 transactions/minute | Better |
| Standard/S2 | 50 | 100 | 500 | 1,470 transactions/minute | Better |
| Standard/S3 | 100 | - | - | - | Better |
| Premium/P1 | 125 | 200 | 2,000 | 98 transactions/second | Best |
| Premium/P2 | 250 | 400 | 4,000 | 192 transactions/second | Best |
| Premium/P4 | 500 | 1,600 | 16,000 | 730 transactions/second | Best |
| Premium/P6 | 1000 | - | - | - | Best |
| Premium/P11 | 1750 | - | - | - | Best |
| Premium/P15 | 4000 | - | - | - | Best |
[1] Database Throughput Unit (DTU): The resources powering each performance level are represented in DTUs. It combines CPU, memory, physical reads, and transaction log writes into a single unit. A performance level with 5 DTUs has five times more power than a performance level with 1 DTU.
[2] Transaction Rate: Transaction rate is the metric produced by the Azure SQL Database Benchmark. It is reported in transactions per unit-of-time, counting all transaction types. In the Azure SQL Database Benchmark, throughput is reported in transaction per unit-of-time, with the time scale being shorter for higher classes of service, as shown below.
[3] Predictability: Consistency of response time is an indication of performance predictability. A database which achieves a more stringent response time constraint delivers more predictable performance. In the Azure SQL Database Benchmark, the response time constraint varies with class of service, with higher classes of service having to satisfy a more stringent response time requirement.
- Windows Azure Table Storage and Windows Azure SQL Database - Compared and Contrasted
- Scott Gu's Blog - Azure: 99.95% SQL Database SLA, 500 GB DB Size, Improved Performance Self-Service Restore, and Business Continuity
- Azure SQL Database introduces new service tiers
- Azure SQL Database Service Tiers and Performance Levels
- Azure SQL Database Business Continuity