Hi

We are in the process of designing and building new silver and gold layers (Star Schema). We will be using Databricks, which is new to the organisation.

• The silver layer will be modelled using classic 3NF, with SCD 7.
• The gold layer is a star schema. We have some facts that will have 9+ dimensions that will require group by for calculating multiple measures.

There is a recommendation by a vendor to use a SHA2 GUID VARCHAR(64) for the primary key of all objects, as DBT is being used for the pipelines. 

There is another proposal to use concatenated strings for the primary keys.

What are the considerations for choosing a primary key data type?

Are there any performance or cost considerations to using a GUID or concatenated string over an Integer data type? 

Thanks

Rob