HP Capacity Advisor 7.2 User Guide

Figure 5 CPU utilization with selected percentiles of utilization
In this example, an allocation of three CPU cores is assumed, rather than the four cores initially
available in this exercise. Even this reduction may not provide the best fit, as the majority of the
work (90%) is completed with less than .5 of one core, and 99% of the work is completed with
one core.
For example, as shown in Figure 5 (page 22), CPU utilization has one peak at 1.7 CPU cores,
with many lower peaks. If you configure your system to always meet the demand of this single
1.7-CPU peak, and you do not adjust the CPU allocation, a significant fraction of the CPU allocation
in this example would go unused most of the time. Depending on your quality of service goals,
you may decide that a different configuration can better use the resources available. Further
experimentation in Capacity Advisor with resource allocations, consolidations, and utilization limits
will help you arrive at the best fit for these workloads.
With Capacity Advisor's visualization and reporting tools, you can make a considered estimate
of server resource utilization using different scenarios and easily refine allocations by tweaking
values in the scenarios. Such estimates can help you minimize overspending for capacity you use
rarely, and maximize utilization of your systems to ensure that your systems have the capacity
needed at the level you require.
Modeling considerations
Capacity Advisor methodology makes certain assumptions in its data analysis. These assumptions
include using an automatically determined CPU performance index, with normalization based on
benchmark data or benchmark data in combination with CPU clock speed in special cases.
Memory is not normalized. Scale factors can be set on the Capacity Advisor Edit Workload and
Move Workload screens. See“Adjusting for platform changes” (page 183) for more information.
On the Edit Scenario screen, you can select either the System or Workload tab to manipulate the
planning scenario and specify which metric representation to view: average, 90th percentile, peak
or max 15-min (maximum 15-minute sustained). The Meter Style can be absolute numbers or in
percentages. By changing the metrics to view, you can explore multiple options for system and
workload deployment. See “ Controlling the data display” (page 67) for more information.
Capacity Advisor commands
In addition to tasks that you can perform from the Capacity Advisor graphical user interface,
Capacity Advisor provides several commands that can be run from the command line on an HP-UX
or Microsoft Windows CMS. For specific information, see “Command reference” (page 193).
22 Features