When creating a knowledge basis on DQS, we can only select SQL Server and Excel from the drop down.
I want to query oracle. To do that, I thought of creating a view on SQL Server that selects from a Oracle linked server but I'm worried about performance.
Lets say my table has 10k rows, I wouldn't mind the impact of running a query against the linked server that returns 10k rows as long as it is run only ONCE.
My question is: does DQS loads all the data it needs to perform the cleansing operation in one go or does it keep submitting queries to the data source (oracle in my example) during the cleansing process?
you'll need to export your Oracle data into a format that can be imported into DQS. It's a one-time data import for a cleansing project. I've made a series of short (<5 min each) screencasts on DQS - this might be helpful as you are getting started.
Also, do install CU1 before you start using DQS as the update improves performance substantially.