Another tricky and rarely discussed reality is that through both tracing Oracle processes or using Oracle's performance views, it is impossible to determine end-to-end response time. Let me say that again-impossible. Don't be fooled by fancy vendors or the promises of tracing and session profiling. Tracing and Oracle's performance views reach from the database server out to as far as the Oracle client process, and then stop. Beyond the Oracle client process, DBAs must use other tools to understand the remaining user experience time.

But it gets even more depressing. By looking at all sessions-that is, at a system-level response-time analysis perspective, the best the performance views can do is tell us about the timing related to the server process. We lose the ability to ascertain the time between the Oracle server process and its associated client process(es).

SQL*Net message from client (SNMFC) is a fascinating and misunderstood wait event. Simply stated, Oracle server processes post this event when they are waiting for a message from their client process so they can do some work. Oracle server processes are either consuming CPU or waiting for something. When a server process has nothing to do, it posts the SNMFC wait event.