The performance analysis reports express most quantified data as
percentages. In most cases, absolute values (for example, actual number
of observations in which execution was in DB2® services)
would not, by themselves, be particularly meaningful. This is because
the total number of samples chosen for an observation session is somewhat
arbitrary. The percentage of activity attributed to a system object,
on the other hand, provides a much better measure of the impact of
that system object on performance. Furthermore, when expressed as
percentages, quantification is likely to remain roughly equal if the
sampling frequency and duration parameters are varied.
In order to effectively interpret the performance analysis reports
it is important that you understand how these percentages are computed.
The formulae vary depending upon what type of system activity is reported.
- CPU Time Percentage
- The percentage expresses
the ratio of attributed CPU to the total CPU time observed. This is
computed by dividing the number of attributed “CPU EXECUTING”
observations by the total number of “CPU EXECUTING” observations
and multiplying that number by 100.
Note that observations of CPU
Waiting and CPU Unavailable are excluded from the calculation. The
objective is to report the relative demand placed on CPU resources
by system objects.
- I/O Activity Percentage of Time
- The percentage reported for I/O activity expresses the ratio
of time attributed I/O operations were active to the total observation
session elapsed time. Consider an example in which 10,000 observations
were made during a 60 second interval. Suppose during 1,500 of these
observations, I/O was found to be active for a file with DDNAME=SYSIN.
15 percent would be reported as the I/O activity percentage of time
attributed to SYSIN.
- Parallel Activity
- Application Performance Analyzer will report information about parallel activity.
Examples of parallel activity are:
I/O activity concurrent CPU execution
is observed.
Concurrent I/O activity is observed for multiple
devices.
Concurrent CPU execution is observed. This is only
possible on a system with multiple CPUs (a multiprocessor).