View previous topic :: View next topic
|
Author |
Message |
fat_frankestein
New User
Joined: 12 Mar 2007 Posts: 2 Location: Hyderabad
|
|
|
|
I need some information regarding the Mainframes usage and billing.
how do we track the total processor time used by all the jobs, submitted by a user, say during a session, so as to say that a particular account has used so many cycles/seconds of the CPU time.
Also, how do we log that data, n how do we extract the same, while billing and reporting |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
Hello,
Usage may be gotten from the SMF records.
If you talk with your system support or operations people, there are probably already processes in place that show usage statistics.
Billing is whatever your organization decides. That answer will come from your organization. Everywhere has their own way to calculate/distribute data center operating costs. |
|
Back to top |
|
|
fat_frankestein
New User
Joined: 12 Mar 2007 Posts: 2 Location: Hyderabad
|
|
|
|
hi.
i knw that the usage is gathered from the SMF.
the usage, i assume, will be logged/recorded in some file. i need to knw how we do that, i mean, what are the fileds that mite be populated( in general, rather than the m/c or the client specific), who, or what system does that job, and how do we extract that data |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
Hello,
As i mentioned earlier, if you talk with your system support or operations people, there are probably already processes in place that show usage statistics.
They can also provide information about how SMF is used at your location.
If you intend to read the actual SMF file(s) you will need to learn about the structure of the SMF records. SMF recording is "automatic". Your site determines what will be captured. Again, the system support people have that information. |
|
Back to top |
|
|
Anuj Dhawan
Superior Member
Joined: 22 Apr 2006 Posts: 6250 Location: Mumbai, India
|
|
|
|
Hi,
For a particular JOB if you want the execution time & the cost..then you can go to JESMSGLG, of that JOB give search on 'unnormalized CPU time'..you will get the time taken by that JOB, near to that time, cost of the JOB will be given.
Perhaps it helps. |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
Hello,
FYI
Quote: |
you will get the time taken by that JOB, near to that time, cost of the JOB will be given |
Only at some sites. Many sites do not present the $cost of the job in the JES output. |
|
Back to top |
|
|
Anuj Dhawan
Superior Member
Joined: 22 Apr 2006 Posts: 6250 Location: Mumbai, India
|
|
|
|
dick scherrer wrote: |
Hello,
FYI
Quote: |
you will get the time taken by that JOB, near to that time, cost of the JOB will be given |
Only at some sites. Many sites do not present the $cost of the job in the JES output. |
Oh..k Thanks for correcting Me. Still why do we have these type of differences at different sites? |
|
Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
|
|
|
|
You're welcome
Not every site has the same approach to billing. Not every site even calculates a $cost for each job run.
What is common to most sites is capturing the amount of cpu and i/o used (SMF). Sometimes you will hear the term "service units", but if one wants to tune, knowing what jobs use the most cpu cycles and what jobs generate the most i/o is a good place to start - this includes resources consumed by database interaction as well.
If the largest resource users can be "tuned" to use less, the $cost (whether calculated and published or not) can be reduced. |
|
Back to top |
|
|
expat
Global Moderator
Joined: 14 Mar 2007 Posts: 8797 Location: Welsh Wales
|
|
|
|
SMF type 30 is probably the place to start. These record what resource is used by which job, STc, TSOid etc. etc.
The RMF records will record all resource consumption, so the RMF total minus the SMF total for any given recording period is the resource used by the operating system which should then be distributed accordingly amongst the billable "users". |
|
Back to top |
|
|
Ravi gaur
New User
Joined: 12 Jul 2005 Posts: 38
|
|
|
|
type30 is correct ..However with the service class TYPE72GO can give you vital information about the cpu uses on the basis of the Service class..like Batch,TSO etc etc. |
|
Back to top |
|
|
|