Re: Performance Data Collection Via Management Central on V5R1

["Alexei Pytel" <pytel@xxxxxxxxxx>]

> I am no so concerned about the performance hit as I am wasting the disk
> space to collect data after 18:00.

If you are not creating database files, *MGTCOL is not taking so much space
- especially if system is idle.

> I guess I am just spoiled by the STRPFRMON launching from the job
scheduler.
> It was so effortless once it was set up and working!  And all the data
was
in one place when done!  :-)

Collector is a different paradigm. I hope it just takes time to get used
to.
PFRMON was expensive and did not work well on very large systems.
PFRMON was supposed to be used on "as needed" basis.
Collector is a low overhead tool and it is supposed to run forever.
In a sense it's even easier - you set it once and forget it. :-)
As for having all the data in one place - creating database files is the
most expensive part. Besides, database files take the most space. So the
decision was made to separate collection work and database creation work.
Some trade-offs had to be made. However, I still think that collector is
not worse than PFRMON - it's just different.

> The other thing that bugs me is why the went
> to the trouble in setting up a parameter to specify how long it runs,
when
> it in fact has no bearing.......

??? Where do you see parameter for how long it runs ? I do not see it on my
Management Central GUI screen. Are we on the same page (<system>
->Configuration abd Service->COllection Services->Properties).
There is a parameter for cycle time, but no parameter for end time.


> I did in fact tell it to create the database files.  The only advantage I
> see so far is that even though the description of the member says In Use,
> you can actually display the data and look at it.

Again, the idea is that you can collect data *all* the time, but do not
spend the expense of creating and storing database files for the time you
are not interested in.
You can create database data *only* for selected periods you want to
analyze.

PS. You can still turn on/off collector from scheduler if you wish.
There is a pair of APIs to start and stop collector (QYPSSTRC and
QYPSENDC).
You can schedule a job with CALL QYPSSTRC ...
and then schedule another job to run CALL QYPSENDC ...

Hope this helps

Alexei Pytel