WLMTK Overview: Using HP-UX WLM Effectively With Your Most Critical Applications

9
Give an instance n CPU shares for each user connection to the instance
In the following example, SLO increases the CPU resources for the instance’s workload based on the
number of user connections. Thus, with five user connections, the workload gets 15%.
Workload: Instance3
Priority: 1
Min CPU: 5%
Max CPU: 90%
Goal: 3% of the CPU resources for each user connection to the instance
Figure 1, based on a graph produced using the optional product HP OpenView Performance
Manager (formerly known as HP PerfView), illustrates this goal with an actual Oracle instance. The
horizontal axis is the wall time, and the vertical axis is CPU resources. The figure also shows the
number of users connected to the database at any given time and the response time (for the response
time, the vertical axis also represents time in seconds).
Figure 1. Shares per metric goal
When the number of users increases (once between 01:30 and 01:45 and again between 02:00
and 02:15), WLM increases the CPU resources accordingly. As a result of the extra CPU resources,
the number of users does not adversely affect the response time. The response time stays roughly
constant (with a slight improvement), providing a consistent level of service despite the increasing
load.