Blog Post

Azure SQL Blog
4 MIN READ

Automatic compute scaling with serverless for Hyperscale in Azure SQL Database

Morgan_Oslake's avatar
Morgan_Oslake
Icon for Microsoft rankMicrosoft
Feb 14, 2023
Optimizing resource allocation to achieve performance goals while controlling costs can be a challenging balance to strike especially for database workloads with complex usage patterns.  Azure SQL Da...
Updated Feb 16, 2023
Version 3.0
azure sql database
Jaap_Bakker's avatar
Jaap_Bakker
Copper Contributor
Mar 30, 2023

Hi Morgan_Oslake 

"In general, serverless auto-scaling occurs without interruption to running queries regardless of their duration."

 

So that works differently for serverless as opposed to manual scaling of vCores in a provisioned hyperscale database?

When we currently scale from e.g. 18 to 24 cores for busier moments based on usage patterns, we see all connections being dropped (only takes 1 to 2 seconds) but that does kill all running queries.

Mostly covered by retry mechanismes, but if serverless does that indeed smoother (e.g. does not drop every connection every time when scaling takes place) then, well, that would be awesome for our use case.

 

Do you happen to know when Hyperscale Serverless will come out of preview mode?

"Serverless for Hyperscale is currently in preview. By using this feature you confirm that you agree that your use of this feature is subject to the preview terms in the agreement under which you obtained Microsoft Azure."