HP-UX Workload Manager User's Guide

WLM quick start: the essentials for using WLM
How to determine a goal for your workload
Chapter 288
Using this configuration, you can directly set an entitlement (allocation)
for a workload using the wlmsend command. By gradually increasing the
workload’s allocation with a series of wlmsend calls, you can determine
how various amounts of CPU resources affect the workload and its
performance with respect to some metric that you may want to use in an
SLO for the workload.
NOTE This configuration file is only for PRM-based configurations. PRM must
be installed on your system. For a similar configuration file that
demonstrates WLM’s ability to migrate cores across partitions, see the
par_manual_allocation.wlm and par_manual_allocation.wlmpar
configuration files in /opt/wlm/examples/wlmconf/ and also included in
Chapter 9, “Example configuration files,” on page 283.
In addition, you can compare this research with similar data for the
other workloads that will run on the system. This comparison gives you
insight into which workloads you can combine (based on their needed
CPU resources) on a single system and still achieve the desired SLOs.
Alternatively, if you cannot give a workload its optimal amount of CPU
resources, you will know what kind of performance to expect with a
smaller allocation.
Once you know how the workload behaves, you can decide more easily
the type of goal, either metric or usage, you want for it. You may even
decide to just allocate the workload a fixed amount of CPU resources or
an amount that varies directly in relation to some metric. For
information on the different methods for getting your workload CPU
resources, see the section “SLO TYPES” in wlm(5).
Similar to the manual_entitlement.wlm configuration, the
/opt/wlm/toolkits/weblogic/config/manual_cpucount.wlm configuration
allows you to adjust a workload group’s CPU allocation with a series of
wlmsend calls. (To see the contents of this configuration file, see
“manual_cpucount.wlm” on page 296.) However, manual_cpucount.wlm
uses a PSET as the basis for a workload group and changes the group’s
allocation by one whole core at a time.