|
View previous topic :: View next topic
|
| Author |
Message |
mvsman
New User
Joined: 09 Oct 2017
Posts: 4
Location: France
|
|
|
|
Hi,
Did anyone ever try to find out top consuming jobs or cics/ddf transactions in TCPIP address space?
I am used to do performance analysis on our mainframe based on MSU consumption exploiting SMF records so I had a look to SMF119 records but I could not find information about MSU consumption....
Any ideas?
Thanks for your help |
|
| Back to top |
|
 |
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8700
Location: Dubuque, Iowa, USA
|
|
|
|
SMF type 70 and type 72 records deal with CPU usage (MSU). You would not find CPU usage in the type 119 records, as far as I know.
MSU is a term associated with the CP or at the lowest level the LPARs defined to the CP. An application CANNOT "consume MSU" because MSU does not exist at that level. The application can use CPU time, along with the other resources, but the application does not use MSU in any way. Furthermore, the ONLY way to change MSU is to replace the hardware -- since MSU is related to the hardware. |
|
| Back to top |
|
|
mvsman
New User
Joined: 09 Oct 2017
Posts: 4
Location: France
|
|
|
|
Thanks for your reply.
| Quote: |
| SMF type 70 and type 72 records deal with CPU usage (MSU). You would not find CPU usage in the type 119 records, as far as I know. |
I know and this is indeed my problem... We do MSU usage analysis with SMF030 (at address space level) and SMF070 (at system level) but I can only get MSU or CPU used by the whole TCPIP address space. And I would like to identify "who" is consuming on TCPIP.
| Quote: |
| An application CANNOT "consume MSU" because MSU does not exist at that level. |
I'm confused. You mean a job cannot actually consume MSU? I have never heard this... Anyway, CPU time will be fine also. |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8700
Location: Dubuque, Iowa, USA
|
|
|
|
A job uses service units (which may be CPU, storage, I/O) but does not "consume" them -- a machine rated at 1000 MSU is always at 1000 MSU no matter how much or how little application jobs are using CPU time. So MSU are not consumed -- they always exist, at the same level, on the machine. Whether or not a job is using MSU doesn't matter at the machine level since the MSU level of the machine will not change until the machine is replaced. A machine rated at 1000 MSU will not be at 970 MSU if a batch job is using 30 MSU; the machine will still be at 1000 MSU.
TCPIP is difficult to break down -- you can have 3270 sessions, FTP client sessions, FTP server sessions, SMTP sessions, etc. -- and there is not necessarily a lot of identifying data in the type 119 records. |
|
| Back to top |
|
|
mvsman
New User
Joined: 09 Oct 2017
Posts: 4
Location: France
|
|
|
|
Ok I understand what you mean. Maybe we have a vocabulary problem here and for sure my terrible english is not helping.
Anyway, if you use all your 1000 MSU for 4 hours, there is little left.. :-)
Thanks for the clarification.
Hopefully, someone else has an idea... |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8700
Location: Dubuque, Iowa, USA
|
|
|
|
| You may have to match the type 119 address space identifier to the type 30 records (address space identifier is in the Identification Section) to be able to see which job(s) are using which resources. The field SMF30WID may help classify work, but you'll have to consider using the interval records as well as the completion records since it is entirely possible for TCP/IP work to span intervals. And I'm not sure if you would be using the subtype 6 records of type 30 or if you can just use the 2, 3, 5 subtypes. |
|
| Back to top |
|
|
mvsman
New User
Joined: 09 Oct 2017
Posts: 4
Location: France
|
|
|
|
Thanks Robert. This is really interesting.
I will investigate that more. |
|
| Back to top |
|
|
baramesh
New User
Joined: 20 May 2008
Posts: 25
Location: bangalore
|
|
|
|
| I have a job that runs for 7 minutes and processed 10 records, and I have another job ran for the same7 minutes but processed 1000 records, how could I determine which is a bad performer? |
|
| Back to top |
|
|
|
|
