What aspects are you interested in?
Teradata is a first-class connector with a full support of everything Teradata has to offer.
It works the same as any other database really - Tableau won't improve the speed of your Teradata server, and what you can do depends on the design and content of the database.
Okay. What connector would give better performance?
Any configurations in Tableau server to improve the performnace and data retrieval from DB?
1 of 1 people found this helpful
What you asked are actually some very big questions. It depends on the size of your Teradata DB and volume of data users would need to pull, the answers can quickly go from one way to another. Here are some general things to consider:
1. Try to locate your Tableau server as close as possible to your Teradata server, same data center or same zone if possible
2. Install drivers per http://onlinehelp.tableau.com/current/pro/desktop/en-us/help.htm#examples_teradata.html
3. Depends on the Teradata server setup and volume of data, if you find live connection is too slow, extracts would be a good choice
4. Try to encourage users to use combination of incremental extracts and full extracts wherever possible: for example, hourly incremental plus daily full, or daily incremental plus weekly full, etc
5. Initial SQL can be used to improve performance by creating temp table, etc. However some inefficient design of initial SQL can create heavy queries to Teradata
6. Query banding allows you pass Tableau server user name to Teradata DB for row level security if you Teradata has user entitlement table but this works for live connection only
7. You should throttle user sessions and have timeout settings from Teradata DB server so self-service Tableau Desktop users will not overload your Teradata DB.
8. You also want to control who should have Teradata direct access by subject areas as well. The recommended approach is to have fewer Teradata DB users who have very good understanding about the data, who are Tableau experts and can do the necessary joins, create all the needed cals in Desktop and then publish the data sources to Tableau server for other workbook publishers to consume. This way you do not have to train too many people about Teradata DB structure, and you can also achieve single source of truth, less load to Teradata DB, scalability, etc.