When mirroring sql database to Fabric, we want to be able to select the column(s) to be used for liquid clustering. This should ideally be available for all mirrored sources such as Azure Sql Database, and also for SQL database in Microsoft Fabric. Liquid clustering can have a huge impact on performance. In our case, we operate a multi-tenant application, with terabytes of data stored in Hyperscale Azure Sql Database. Virtually all tables have a tenant_id column. To offer in-app analyics, we run some analytical queries. Since each tenant is isolated, we always have a where clause on the tenant_id column (WHERE tenant_id = xxx). Without clustering, the sql analytics endpoint of mirrored database must scan the data of all thousands of tenants. Clustering would ensure that all irrelevant files for other tenants can be skipped. This is the one reason why we cannot leverage Fabric for now.
... View more