I have had this issue and there isn't really a "fix" there is workarounds. The best is to pull copies of the data local, using something like SSIS because it can do Fast Bulk Load.. Another option is a stored procedure on the remote side that does as much pre-processing as possible and then returns a sub-set back that is held in a temp table, similar to the first option. Another option is to re-architect the process to make only discrete calls remotely and only work with remote data when it ABSOLUTELY needs to. My last thought is join criteria, it *seems* that when the joins are primarily on numeric (int, numeric, etc.) columns the optimizer does a better job and it performs better. I can't substantiate that but that has been my annecdotal experience..

CEWII