HP-UX Workload Manager User's Guide

Configuring WLM
Example configuration
Chapter 5 233
For more information on how to use the cntl_margin tunable, see the
white paper “Tuning HP-UX Workload Manager” at
/opt/wlm/share/doc/howto/tuning.html.
Releasing cores properly (optional)
By default, the cntl_base_previous_req tunable is set to 1, which can
be beneficial when you are using the WLM global arbiter (wlmpard) to
manage virtual partitions or nPartitions or when your WLM
configuration has at least one PSET- based workload group with an SLO.
WLM creates a controller for each slo structure that has a goal
statement. Controllers make requests to the WLM arbiter for shares
allocations. Setting cntl_base_previous_req=0 can result in some
controllers requesting CPU resources more aggressively; however, the
distribution of cores may become unfair due to artificially high resource
requests.
Use this optional tunable in global, metric-specific, and
metric/SLO-specific tune structures
Example configuration
The following is an example configuration consisting of two workload
groups and four SLOs. A discussion of the example is presented after the
example.
# Define the workload groups
prm {
groups = finance:2, sales:3;
apps = finance: /opt/fin_app/count_money, sales: /opt/sales/do_ebiz;
}
# This is an SLO for the finance group.
slo finance_query {
pri = 1;
mincpu = 20;
maxcpu = 50;
entity = PRM group finance;
goal = metric fin_app.query.resp_time < 2.0;
condition = Mon - Fri;