ORA-02393: exceeded call limit on CPU usage -- Concept Understanding is req

Question

In our System CPU_PER_CALL is set to 1.5 Hours for Reporting Users.
I can see some query runs for 10 hours-15 hours and complete successfully and some queries fail exactly after 1.5 hours.
I want to understand what does CPU_PER_CALL Means. On what basis it calculates CPU_PER_CALL ( Fetch , Execute , parse). How a query is calculating time ?
With the same profile options some queries run for 10 hours but some queries fail after 1.5 hours.
Regards
Sourabh Gupta

Answer 1

The short answer is that different queries wait on different sorts of events. Let's assume that the only 2 wait events in the world are waits for CPU and waits for I/O (there are many other types of waits but most reporting queries will primarily be waiting for these two resources). If you have a query that runs for 15 hours but spends 14.5 hours waiting on I/O and only 0.5 hours on the CPU doing comparisons and/or calculations, the CPU usage for that query is only 0.5 hours. Another query might run for 1.51 hours and do 0.01 hours of I/O and spend 1.5 hours on the CPU calculating various aggregate values for that data. The second query would use 1.5 hours of CPU (and thus exceed your CPU_PER_CALL) while the first query would only use a third as much CPU.
Oracle profiles allow you to specify a number of different limits so that you can specify limits on CPU usage (CPU_PER_CALL/ CPU_PER_SESSION) or I/O usage (LOGICAL_READS_PER_CALL/ LOGICAL_READS_PER_SESSION) or a combination of the two (COMPOSITE_LIMIT).
Justin