Unanswered: who does fully understand "throtteling" in db2 luw?
What I do know:
- at instance level we've got the util_impact_lim cfg paramater
- the command syntax has extra parameters to overrule this instance paramter (util_impact_priority)
- you can overrule an already running utility with "set util_impact_priority"
What I do not know:
- is there a hierarchy involved? Suppose I have 3 instances with more databases and i've got the dbm cfg set to 30. Next I start a backup with util_impact_priority = 80. Will the backup get 80% of the 30% assigned to that instance or will be backup get 80% of the total system resources?
- is the throttle looking at the current workload? Suppose I have a dedicated server (no other workload running) and I start a backup with trottle of 50%. Will it run as fast a when I start untrottled (because it does not have to share any other resources) or does is run slower and the system itself is not under full load because of this parameter?
Suppose I have 3 instances with more databases and i've got the dbm cfg set to 30. Next I start a backup with util_impact_priority = 80. Will the backup get 80% of the 30% assigned to that instance or will be backup get 80% of the total system resources?
As far as I know, it's neither. In your example, if the backup is the only throttled utility, it will get all of the 30% of the server resources as defined by UTIL_IMPACT_LIM. If, on the other hand, you have a concurrent REORG with UTIL_IMPACT_PRIORITY of 40, the backup will get 30*80/(80+40) = 20% of the server resources, while REORG will get the remaining 10%.
Please evaluate my assumption:
- util_impact_lim dbm cfg is set to a value
- I have an idle system and I must perform an online bakcup as fast as possible
- I must specify UTIL_IMPACT_LIM = 100 in my backup command to overrule the dbm cfg
- In this situation an online backup executes unthrottled and is as fast as an offline backup (all our restore scripts are based on an online image + rollforward) and all server-resources are used
(or have it your way: I am too stupid to understand)
I didn't say that, did I?
Originally Posted by dr_te_z
Then what does the DBM CFG do???
UTIL_IMPACT_LIM sets the limit of server resources for all _throttled_ utilities. You make utilities _throttled_ by specifying UTIL_IMPACT_PRIORITY. If you don't specify UTIL_IMPACT_PRIORITY, the utility runs unthrottled and as such it is not subject to the instance-level limit.
I re-read you link and I looked up the exact meaning of the word "irrespective" and now I understand that there is an "and" relationship between the DBM CFG setting and the "util_impact_prio" parameter in the commands. I was under the impression that the DBM CFG set a default for all ultilities and the "util_impact_prio" was used to overrule that.
Thanks for your help of making that clear.
Next weekend will be our conversion weekend from 9.5 to 9.7 and we are not allowed to use too much "downtime" so the restore & backup's MUST go at full speed.