Automatic Concurrency Scaling is a feature offered by several cloud service providers such as data warehouses. It enables automatic addition and removal of computing capacity based on demand from thousands of concurrent users.
This automatic scaling is based on so-called "multi-cluster" data warehouses, made up of several clusters executing requests. Users can configure the minimum and the maximum number of clusters allocated to each warehouse.
The automatic scaling mode, therefore, allows cluster capacity to be added automatically in the event of an increase in concurrent read requests. This way, the user receives more computing power when needed.
Once this additional capacity is no longer needed, it is automatically removed. The user, therefore, does not need to manually manage scaling.
If the number of requests linked to a scaling queue exceeds the configured maximum competition, eligible requests are transferred to another cluster. This causes:
Thanks to this automated scaling, users can benefit from:
Automatic Concurrency Scaling also addresses one of the main weaknesses of traditional data warehouses, where computing resources are fixed and do not adapt according to the number of queries.
This advantage is based on the fact that cloud-based data warehouses allow the separation of computation and storage. In addition, scaling is instantaneous, so users only pay for the resources they use. The savings can be significant.
This feature is ideal for high query volumes, as it will significantly increase performance. Users can therefore cope with the increase in concurrent requests and load changes that occur throughout the day.
This automatic scaling mode is found in many cloud data warehouse providers. However, it is an innovation that has not yet been democratized to all vendors. For the moment, this mode is mainly offered on Amazon Redshift and on the Snowflake platform, considered as the two leaders in this field.