While Monitoring CPU usage from DBQL for each user, grouped by day/hour/statement type.
we find that [CheckPoint Loading] statement type consumes Most of CPU for any user uses utilities , that refer to more than 90% of CPU utilization accross the day.
We are using a query like this one:
sum(AMPCPUTimeNorm ) SumCpu
where logdate between '2013-01-27' and current_date
and username in ......
group by 1,2,3,4;
The strange thing is we do not set 'Checkpoint' for any of our utuilites parameter files;
so it should use the defauk value which AFAIK is 100K rows, and we encounter the problem for users loading tables with 500K rows.
i'm worried that the CPU accounted for [CheckPoint Loading] in DBQL is actually total CPU used in the loading job itself not only due to the checkpoint.
You are correct. The CPU attributed to [CheckPoint Loading] includes load processing as well as the checkpoint itself.
Note that there is a checkpoint at the end of acquisition phase, even if you disable intermediate checkpoints.
Most of CPU in Our system Spent there, i'm thinking of using NOPI which May help to decrease CPU spent there (which is our main concern) ; any other recommendation ?