OSI/AS Configuration and Management Manual

Performance Considerations
OSI/AS Configuration and Management Manual—424119-001
6-15
Corrective Actions
should analyze what is happening in the system, determine the causes, and then attempt
to treat the problems—not the symptoms.
The various performance indicators that you can monitor describe different aspects of
the behavior of a subsystem. However, these indicators are generally not independent of
each other; they are interrelated. Changing one factor or attribute, or making haphazard
changes to attributes doesn’t always improve the functioning of your system. The net
effect may be to make the situation worse.
The overall goal is to avoid the overutilization of your subsystem resources. General
guidelines suggest spreading the workload evenly across all the hardware and software
resources that make up the subsystem. To reduce the overhead, you can add
communications lines, disks, CPUs, and other hardware. You can restrict throughput by
controlling the number of processes on the subsystem and by making the application
programs more efficient. If queueing delays on a communications line are excessive,
you can increase protocol window sizes or increase line speed. If you cant add more
hardware, you can reduce the less-important activities during peak periods.
Any proposed changes should be tested before introducing them into the production
system. If you cant test them, you should carefully consider the implications of the
changes and monitor the subsystem closely after the changes are implemented. Each
change you make to the system can affect a number of different system resources.
Finally, keep in mind that making changes to any one of the factors discussed in this
section may have only a minor impact on total system performance. However, making
changes to a number of these factors together can result in improved levels of
performance.