Dispatching Priority of Db2 Monitors

Max Scarpa

Dispatching Priority of Db2 Monitors
Hi Tekin

It's a general rule (stated by many papers) if you're using WLM then
monitors (all monitors) must have a SERVICE CLASS (which is then
'converted' in the proper DP by WLM - WLM works in this way) equal to the
service class of the monitored address space. In WLM hierarchy monitors
(and even some 'tools' using DB2 traces) should be a kind of 'supervisors'
(lato sensu) so they have to stay in a equal or higher position and this is
more important if you're using monitors interacting with DB2 (for instance)
using 'expensive' traces or more than one trace generating many data.

Don't forget that if you have all monitors in the same, lower service class
and your machine needs CPU to reach the goals of other address spaces in
higher service classes then ALL the monitors are delayed, causing even more
delays in monitored AS - a classic in a 100% CPU. If you've monitors in a
service class with a goal lower than DB2 it's like to have a hundred-metre
runner with a iron ball hooked to her/his peg. This may be more evident in
online (you see spikes in elapsed time when the monitor is not served) and
in long running batch jobs. You can see this when monitors service class
has a PI > 1.

Here we have a service class called 'Monitor' and it's higher than the
service class 'STCHIGH' where DB2 resides
(it's just below SYSSTC service class).

HTH
Max Scarpa
The usual, useless, hyper-expensive ,boring DB2 sysprog