HP-UX Workload Manager Toolkits User's Guide

HP-UX WLM Oracle Database Toolkit: Providing Database Metrics to WLM
How do I use the metrics?
Chapter 254
To implement your own time-based goals from SQL statements:
Step 1. Identify SQL statements that perform the transaction to be timed.
Step 2. Try the SQL statements manually using SQL*Plus.
NOTE Oracle is very efficient at caching data. When you are getting a walltime
by hand for a baseline, run the query twice using the option
--iterations=2 to wlmoradc. The first iteration encaches the data, and
the second iteration represents the query with the data encached,
providing a more realistic execution time.
Whether data comes from a disk or cache memory is not important.
WLM and ODBTK are only concerned with the walltime needed to
execute the SQL statements. Either way, if your data stays encached or
if the system is so busy that data is quickly moved out of cache memory,
your walltimes represent the typical times for queries on the system.
Step 3. Place the working SQL, along with any desired wlmoradc options, in an
wlmoradc configuration file, say times.oradc.
Step 4. Invoke wlmoradc on the command line:
# wlmoradc --configfile times.oradc
If there are problems, add the option --debug 2 to see the SQL input
and output and determine the causes:
# wlmoradc --configfile times.oradc --debug 2
Step 5. Invoke wlmoradc from a WLM configuration file
a. Specify your SLO.
Here, the goal is to keep the transaction time under 300, with the
time units determined by the data collector used.
slo Ora_1_slo {
pri = 1;
mincpu = 15;
maxcpu = 75
goal = metric O1_transaction01_time < 300;
entity = PRM group Ora_grp_1;
}