I was looking into one of the STROBE result for one of the batch job for which the elapsed time was on very high end.Finally from STRETCH time our tuning group concluded that probably we might need additional CPU power to get the ET reduced.We thought of initiators also along with runnning job with some other class.
I saw the stretch time in STrobe report.
I thought of getting all these type of situation where the stretch time is more for jobs I mean to first identify all such types of jobs where job is ready to run but need additional resouces like situations.
I am wondering what type of dataset or SMF record I should target for starting with.I think would have to use some SAS code to report further.
Please move this post to any appropriate place if this is not correct.
Joined: 06 Jun 2008 Posts: 8231 Location: Dubuque, Iowa, USA
IIRC from earlier days, stretch time merely means the address space is not executing. That could be due to many factors.
The RMF report on CPU time includes data about address spaces ready to execute but not able to -- search the RMF manual for details. If I get a chance tomorrow, Ill look up the exact report and line that has this.
Using the following code on pur production Lpar trend dataset I am able to report value for JELAPSTM for jobs but just thinking if elapsed time would upto some extent would speak for Stretch time also.Beyond this point I really need to research and experiment to conclude something on the Stretch time report which I would need.
DATE = DATEPART(JINITIME);
FORMAT DATE DATE9.;
TIME = TIMEPART(JINITIME);FORMAT TIME TIME12.;
IF TYPETASK = 'JOB';
IF JOB =:'K' ;
PROC SORT DATA=JOBS;BY DESCENDING DATE;
PROC PRINT DATA=JOBS;
VAR DATE TIME JOB CPUTM JELAPSTM CONDCODE
EXCPDASD EXCPTAPE TAPEDRVS TAPNMNTS TAPSMNTS;