Provisioned Throughput
Ideal for always-on production usage
Throughput can be defined at the database or container level
Requires 10 RU’s per GB of storage
When defined at the Database level it is called as Shared Throughput
If specified at the Database level then all the containers in the Database (up to 25 containers) has to share the throughput
At the container level it is called as Dedicated Throughput
This approach is recommend to achieve predictable performance
The container throughput is defined by partition key, size of data and other factors
Once RU’s are consumed subsequent requests are Rate Limited
By default manual scaling is required to acquire more RU’s
We can configure autoscaling where for throughput we provide a maximum value
The minimum is calculated as 10% of the maximum value
Serverless Throughput
Pay only for the RU’s consumed and the storage used
Ideal for development and non-frequent workloads
Supports a maximum of 5000 RU’s