This isn't a Perfkit answer, but you could create an accounting record every 10 minutes. It would have the CPU usage data for the previous 10 minute interval. If it is only a few virtual machines, then instead of running ACNT ALL for all machines, just run 'ACNT userid' for each user that you want the data for. This assumes that you have enough space on the DISKACNT 191 disk and that creating extra records won't cause any problems with any existing use you have for the records.
Of course, if you need more information than is provided by accounting records (which is usually the case!) then you'll need Perfkit. On Mon, Jul 26, 2010 at 11:00 AM, Michel Beaulieu <[email protected]>wrote: > Good Day, > > Thanks for the feedback from both of you. > > You gave me useful tips that I can act upon. > > This week I will experiment with the following approach: > > The settings will be changed to have extended summary data saved > to "systemid FCXSUMMnn" on a 10 minutes interval. > > I will copy and rename these files regularly on another minidisk. > That will become some kind of extended archive. > I want to stay away from processing raw monitor data as much as possible. > Later I will be able to get other fields information besides CPU seconds > currently. > > Then, I will see what I can extract using REXX and/or PIPELINE > to extract information from 'FC41' records (by userid) and 'FC42' records > (by user class). > > I already have the REXX code to convert simple floating point fields. > (C2F). > > I will have to find or create some code to convert the TOD field. (That > should not be too hard) > > I will keep you and this list informed on my progress. > > Regards, > > > Michel Beaulieu > Montreal, Canada > |*| > -- Bruce Hayden z/VM and Linux on System z ATS IBM, Endicott, NY
