I have, in the past, created views on the local server that contained the required remote data. Although you don't get quite the performance gain from copying the data down from the remote server, it allows you to maintain a granular security structure at the remote location, and you do acquire some performance gains over direct access. From a security standpoint, you only expose that data that needs to be exposed. One "gotchya" on this arrangement is that you now expose your local architecture to vulnerabilities dependent on the schema consistancy of the remote server(s). If someone changes the remote table schemas without notification / propegation of the changes, things on the local server . . . ah . . break.