HP-UX Workload Manager User's Guide

Configuring WLM
Example configuration
Chapter 5 235
This configuration file specifies four SLOs. The finance_query SLO is
active Monday through Friday. Its goal is that the metric
fin_app.query.resp_time, which is provided by the executable
/opt/fin_app/finance_collector, must always be less than 2.0. This goal is
priority 1, the highest priority. The controller for this goal determines the
CPU entitlement (allocation) needed to achieve the goal, but is bounded
by the minimum and maximum CPU resource requests, which are 20%
and 50%.
Additionally, a stretch goal is defined for this workload. The stretch goal
is the SLO named finance_query_stretch, which is also only
applicable Monday through Friday. The stretch goal is of lower priority
(pri = 5). The stretch goal is that the metric fin_app.query.resp_time be
less than 1.0. The controller for the stretch goal is bounded by the same
minimum CPU allocation, but has a higher maximum.
The SLO named finance_query_weekend applies to this workload
group as well. This SLO does not specify a performance goal, but
requests a CPU allocation of 5% for weekends.
The sales_query SLO is active at all times. Its goal is that the metric
sales_app.resp_time be less than 10.0. This metric is provided by the
data collector /opt/sales_app/monitor. This SLO specifies that the
controller for this SLO never requests a CPU allocation of less than 40%,
nor more than 80%. This goal is priority 1.
The global tunable wlm_interval is set to 30, thus the WLM daemon
checks for new performance data every 30 seconds.
The finance_query and finance_query_stretch SLOs both use the
metric fin_app.query.resp_time. The tune structure for that metric
specifies the settings of the tunables used by those SLOs. The argv for
the process used to collect the metric fin_app.qurey.resp_time is given by:
coll_arg = /opt/fin_app/finance_collector -a 123 -v;
For the controllers that use this metric, the value of the proportional
constant is set to 1.0. All other tunables are set according to the default
values listed in the master tunables file.
The sales_query SLO uses the metric sales_app.resp_time. The tune
structure for this metric specifies that when the daemon starts the data
collector to collect this metric, it uses the coll_argv string
/opt/sales_app/monitor -threshold 2 -freq 30. Also, this tune
structure specifies the proportional constant used by the controllers for