When you say that 80% of the wait time is sequential reads, is that counting CPU time in the total (i.e. sequential reads/(total waits + CPU) = 80%)? Since you haven't yet run a 10046 trace, I'm guessing that you got your figures from a statspack report. Is your DB 8i or 9i? How much parsing are you doing? Many will tell you not to use aggregate numbers (e.g., statspack), and Cary's approach is the best one for solving specific problems. His paper on Performance Management Myths and Facts (http://www.hotsos.com/downloads/visitor/00000024.pdf) has an excellent explanation of the risks of more CPUs. But you can get a lot of valuable information from your statspack about whether more CPU will solve your problem or add to a bottleneck. --Terry ----- Original Message ----- List - We are trying to guess the benefit from doubling the CPUs on a server, from 2 to 4. Will this increase the performance or just result in an I/O bottleneck? This server supports a production application and the season of heavy usage is approaching. A development team is writing the replacement application, so we are just looking at something to increase the capacity for this season. In addition to Oracle, a piece of the application runs on the server. During the production season, the CPU remains pegged at 100 percent most of the 24-hour cycle. But if you look at the overall Oracle wait times, db file sequential read is about 80% of the wait. I have read Cary's book and suggested running a 10046 trace, but would like input on the best approach. Thanks. -- //www.freelists.org/webpage/oracle-l -- //www.freelists.org/webpage/oracle-l