Yep, 1 major performance issue we have occasionally. SAS allows you to treat an oracle table as a SAS dataset, allowing you to access it using a datastep rather then using a Proc SQL step and a "Connect to Oracle" statement.
However, it is worth bearing in mind that if you do this, the process extracts the entrire table before processing it, ignoring all where statements until afterwards. This means that if you are extracting say a small number of accounts from a view which touches several tables, you can find it extremely slow going as the entire table will be extracted, then the where statement applied next. In these cases it is best to use the Proc SQL step rather than a datastep. It can make a difference of about 20 minutes on some of our views.