"Bryan O'Halloran" <[log in to unmask]> wrote:
 
>We run our own home grown data collection program using the MI.  We collect
>"heaps" of statistics and the job only uses about 1,000 CPU seconds each 24
>hours.
>
>It seems fairly efficient for the amount of work it does.
>
 
You have missed the entire point of this thread.  I suggest you go
back and reread it.  You have used about 1000 CPU seconds that you
know about and countless more that you have no visibility into.  Most
of the time when people run these background jobs to collect
statistics from the MI, they have effectively replaced their system
with a smaller one, one or two models down the scale.  I hope they
enjoy those graphs, because they are some of the most expensive graphs
I can imagine.
 
/Steve Cooper