Recently, I came across a question regarding performance of Big SQL federated table in Teradata and a Big SQL tuning expert's answer. It's worth sharing and may be it can help you as well. Note: This performance tip is specific to the system and may not be applicable to all environment. Should you have further questions/concerns, please consult with IBM Consult/Support team.
Analysis from an Expert:
Big SQL optimizer re-writes count(*) queries to count_big(*), since this function is not mapped by default that would mean the query would not be pushed down, instead all the rows would be fetched and the count would occur on bigsql side. On small tables you would not notice but on larger tables you would definitely see some performance deterioration. You can check that this is what is happening by getting the plan for the query and checking the RMTQTXT on the ship operation.
Possible way of fixing this issue is by creating a function mapping for count(*) as below example,
CREATE FUNCTION MAPPING count_big_mapping FOR SYSIBM.COUNT_BIG() SERVER tdremoteprod_pdcrt OPTIONS (REMOTE_NAME 'COUNT(*)');
After creating the wrapper, the results were returned in less than a second.