Hi,

Ordered in estimated possible gain:
use of indices/rushmore for the views (fits with data size dependancy, also check set collate)
"blocking" issues of rushmore by indices with low discrimination (fits with data size dependancy)
"use" is sometimes slow in some OS combinations if the table is already opened by another machine
"use" is sometimes slow, especially if set("deleted")="ON" (fits with data size dependancy)
put read-only lookup tables to the workstations (fits partially with data size dependancy)
pathing issues (doesn't fit with data size dependancy)
locking issues in the dbc
exe / dll's on the server (if made read-only, otherwise higher up!)

Based on those numbers you should have no problem if the tables are indexed correctly and the views return only highly filtered sets. If you return a couple of hundred MB's, even client/server approach won't help too much.

HTH

thomas