GPA Manual for G-series RVUs

Guardian Performance Analyzer (GPA) Manual – (544977-001) Page 69 of 106
From Example 4-13 you can see that on \NODEB, the least utilized processor, CPU 3,
is out of balance. Since the percentage busy times for the other three processors fall
within the target utilization for the system (discussed later in this section), GPA
considers these processors to be in balance.
The next chart in this section of the report identifies each of the processors on the
node with respect to number and type. In addition, the chart shows explicitly how the
values for a number of performance parameters are distributed among each of the
processors. It thus indicates which are the problem processors on the node. Here you
can also see more precisely than in the Processor Load Balance Chart what the
percentage busy time is for each processor.
The Processor Performance Chart also shows how many disk processes (DISK
PRIME) are primaried on each CPU. Notice in our examples how unevenly the primary
disk processes are distributed among the CPUs, which accounts partly for the load
imbalance condition on these nodes. (One GPA function is to determine how such an
imbalance could be corrected by redistributing the disk processes among the CPUs.)
Note that for \NODEC (Example 4-19), the imbalance is severe enough that GPA
makes a recommendation for primary disk process changes and generates “After
Primary Changes” sections in its standard report.
Other important parameter values shown in this chart are:
• The page swap rate (SWAP RATE). As mentioned previously under System
Performance Summary, a page swap rate greater than 1 indicates a memory
shortage or overutilization of the CPU.
• The number of pages of memory owned by the memory manager. Refer to your
SYSGEN output and note that at the end of each processor section are listed the
number of pages locked and the minimum pages. Before any processes are added
to a processor, the number of memory manager pages should be reduced by the
minimum pages value and the swap rate of the processor should be reviewed. If
either the current swap rate is greater than 0.25 or the adjusted page number
value is less than about 200, no additional processes should be added without
adding physical memory.
• The message rate (MSG RATE) and the dispatch rate (DISP/SEC). Generally,
there are at least two dispatches associated with either a send or a receive
message.
The last line item on the chart gives the halt impact for each processor, the effect on
system performance if the processor were to fail. Example 4-12 shows that failure of
any one of the CPUs on \NODEA might have serious consequences. The term
MEMORY here means that failure of the CPU could result in a memory shortage in
one or more of the other processors which, in turn, could substantially degrade the
system’s performance.