I am in the process of designing a SaaS system that will be running in Azure and utilizing SQL Server. The application will leverage an Azure Elastic Database Pool to support scaling out via Sharding. The plan is to run a Shared-Sharded model and leverage a Shard Map to map the Tenant Id to the correct database.

I've been reading the Azure documentation and have successfully setup an Elastic DB Pool across a couple of my test databases in my DEV environment. So I understand how to do implement and in theory how this will work.

What I am unclear on is whether it is best practice to create each Azure SQL Database (Shard) on it's own Azure SQL Server or just to run one Server with multiple databases. I'm trying to work out whether there is potentially any performance bottleneck if I had 30 databases on one server vs. 30 servers with one database each. In either circumstance I am assuming the 30 databases are all part of a single Elastic Pool.

Can anyone help me understand conceptually how the resource are allocated? All the documentation talks about the resource being allocated to the database and the server is never mentioned so I am assuming this is somewhat irrelevant and is simply a container in the this context.

This question is driven by the bulk of my experience being based on a world of physical SQL servers where there is a limit to how many databases you can resource on each server.