Ad Hoc Query Performance in Distributed Environment

     In an attempt to avoid replication of large quantities of data, we have found ourselves in a couple of situations in which we've set up database links for use in a data warehouse application. As you might guess, performance has been inconsistent. Some queries do well, others run for an extremely long period of time.

     Questions:

1. Has anything changed between Oracle 7.0 and 7.1 to significantly improve the performance of distributed queries (those accessing tables on the local server and one remote server)? between SQL*Net 1 and 2?
2. Are there books/articles that explain in some detail what the optimizer does with distributed queries? The Concepts Manual says very little, the main point being that "ORACLE decomposes the statement into individual fragments, each of which access tables on a single database." (p. 13-66)
3. Are there books/articles (or experiences) that illustrate the variables which have a significant effect on optimizing the performance of distributed queries? I'm thinking about parameters of Oracle, SQL*Net, the underlying network [we use TCP/IP], the physical layer, etc... In other words, given the problems of ad hoc distributed queries, what possible accomodations could a system administrator make to improve the situation?

     Thanks for any thoughts you might have about the above questions (or related topics)... BTW, my gut feeling is that we will end up doing some replicating, but I'd like to work out some principles/guidelines that tell me when I need it and when I can aovid it.

                              Dave McLaughlin
                              Information Technology Division
                              University of Michigan