Add a data flow task with separate data flow source tasks for Server A and Server B Then attach the results using the appropriate data stream conversion task.
As an example, this data stream takes the Flat File Source and OLEDB Source tasks, sorts the results, then uses the Merge Join task for the results. It looks like your implementation will require two OLEDB Sources or (ODBC, ADO NET, etc.).
I like this method on a linked server or OPENROWSET because you do not need to configure a linked server or enable Adhoc Distributed Queries in your SQL Server data sources.

Bryan source share