Service tiers and performance levels

SQL Database service tiers differ based on performance levels. Within each service tier, performance levels are designed to have a database operate as if it is running in its own machine, isolated from other databases and services. SQL Database accomplishes this by using a resource governor for certain resources and hard resource limits for other resources.

Each database can have its own service tier with its own limits, independent of any other database. This option is best for databases with a consistent load. You also can place multiple databases in an elastic pool where resources can be shared among databases and minimum and maximum resource thresholds can be set for each database in the pool.

The resource governor enforces a maximum amount of resources for either a single database or an elastic pool of databases. If the aggregated resource utilization reaches the maximum available CPU, memory, log I/O, and data I/O resources assigned to the database, the resource governor will queue queries in execution and assign resources to the queued queries as resources become available.
To assist you in choosing the amount of resources you need for your database, Microsoft developed a unit of measure called a Database Transaction Unit (DTU). To develop this benchmark, Microsoft took a set of operations that are typical for an online transaction processing (OLTP) request and then measured how many transactions could be completed per second under fully loaded conditions within each performance level.

If you are migrating an existing SQL Server database, you can use a third-party tool, the Azure SQL Database DTU Calculator, to get an estimate of the performance level and service tier your database might require in SQL Database.


DTU resource limits
Single database: A single database can have between 5 and 1,750 DTUs. A Basic database has 5 DTUs, which means it can complete 5 “typical” transactions per second, while a Premium P11 database supports 1,750 DTUs.


Elastic pool database: A database can be a member of an elastic pool with between 100 and 1,500 DTUs. All databases within a pool share a common set of resources.


Other resource limits
SQL Database enforces limits on other resources by denying all new requests when limits are reached. These limits include the following:
 Maximum database size: Between 5 GB and 1 terabyte (TB)
 Max in-memory OLTP storage: Between not available and 10 GB
 Max concurrent requests: Between 30 and 2,400
 Max concurrent logins: Between 30 and 2,400
 Max sessions: Between 300 and 32,000
 Max databases using automated export: 10

Source of Information : Migrating SQL Server Databases to Azure