SMF 30 and subtype V.

[LGhus]

Hi,

In one project in other company I have created a report on SMF records with the help of other guys. This report included SMF 30 subtypes 4, 5 & V. At this new project when I am trying to create similar report but now I am getting confused if subtype V even existed as I could only found the description about subtype 6.

[Robert Sample]

According to the manual, as of z/OS 2.1 the ONLY subtypes that exist for SMF record type 30 are 1 through 6. The subtype V is something the people in the other company came up with, and unless you work with someone from that company to get the method used to create it, you have little to no chance of reproducing it. Use the standard record subtypes to accomplish your goal.

[LGhus]

Thank you.

If I need to look for STC (started task) SMF 30 subtype 4 should help me? I looked at notes in this link http://www.pacsys.com/smf/smf30_v1r11.htm.

[Robert Sample]

If you are looking at started tasks, subtype 4 would be the LAST place to look for information -- subtype 4 is step termination and a started task generally does not terminate. For started tasks, subtype 2 (interval accounting) would be where most of the information should be coming from; subtype 6 (system address space) may be necessary depending upon the started task, and subtype 5 (job termination) may be used after an IPL (since the started task has to end before the IPL occurs). Subtype 1 would be used immediately after an IPL when the started task is started. The subtype 2 data is usually, but not always, collected at a site and the interval is controlled by the SYS1.PARMLIB member SMFPRMxx.

[LGhus]

Thanks Robert. I did find this information on 30_V:

WHAT IS SMF TYPE 30_V ?
In very general terms, the RMF records mentioned above measure usage from the perspective of system components ie CPU, Memory, DASD, etc. The Type 30 records on the other hand measure activity from a workload perspective. Sub-types of the Type 30 records are written by SMF at certain stages during the execution of a workload event. Processing of the Type 30 records, any SMF record in fact, is usually done using The SAS System software to produce tables for viewing and subsequent processing.
A number of ‘off the shelf’ applications consisting of a suite of SAS programs to extract details from SMF are available. One such product, Merrill’s Expanded Guide to Computer Performance Evaluation (MXG), is one of the most popular products used to create a series of SAS tables known as the Performance DataBases(PDBs).
The Type 30_V records are built by MXG into a SAS table named SMFINTRV. As the name suggests, these are records written at the end of a pre-determined interval period for every unit of workload being executed on the system. The length of the interval period can be varied, but tends to be set at 15 minutes. They are a good source of information to assist systems programmers with the resolution of problems because they list all workloads being executed on the system at the time of a problem."

I am now confused if it is subtype 5 (roman 5 = V)

[LGhus]

I could again spot the link which has the above definition ... are they essentially talking about subtype 5.

http://www.cmgaus.org/cmga_web_root/pro ... stle96.pdf

I also looked at http://www.pacsys.com/smf/smf30_v1r11.htm for the definition subtype 5 but I am not able to comphehend if they both are talking about same things.

[Robert Sample]

I've worked quite a bit with MXG in the past, and it would have helped in your original post if you'd mentioned that you were using MXG to work with the SMF records. 30_V records in MXG are based on subtype 2 -- not 5 -- as they reflect the interval data written to SMF. I'm sure Barry Merrill had a reason to call then 30_V but I don't know what it is. The SMFINTRV data in the MXG PDB is valuable for capacity planning and performance management as it provides resource usage by WLM service class.

[LGhus]

Thanks Robert.

So if we are using MXG, it might provide us with its own names of records for the existing SMF types from IBM?

[Robert Sample]

MXG does NOT use the pure SMF records as written to the SMF data set -- MXG combines some and otherwise manipulates the data so when you are using MXG you really are not dealing with SMF as such. For example, if you are using a third-party product that creates SMF records (with record ids in the 128 - 255 range), you won't see anything in the MXG PDB about any of these records unless and until you do the customization in MXG to add them to the PDB. The records are cut and present in the SMF data sets but not the MXG PDB.