Mainframe, MVS and zOS Discussion

Hi,

I'm on a performance Testing assignment and trying to get the response time for CICS Transactions from SMF 110 record type. Using DFHSTUP got the readable format from SMF records. However, the transaction start time and transaction end time is not matching with the real system time. I'm not sure how should I proceed from here, any help in the analysis of SMF 110 record layout would be great.

However, the transaction start time and transaction end time is not matching with the real system time

What does this mean? What are you considering the "real system time"? SMF data may be difficult to interpret correctly, but it is rarely in error since the system is maintaining it.

Hi Robert,

We've given a SAS Graph which has the transaction start and end time. We also looked at SMF 110 records to ge these values - we took time from SAS graphs as standard system time. These two timings are different.

Depending upon which specific field you are looking at, SMF records times as a binary-coded decimal, number of milliseconds since midnight, or number of clock tics (there may be other formats I'm not recalling right offhand). Furthermore, SMF data may be recorded using UMT as opposed to local time. So you may have multiple conversion steps required to take the SMF time value and make it something meaningful.

Thanks Robert.

For a novice in SMF, what is the best resource for SMF/RMF yuou'd suggest?

If your site has MXG, the MXG SOURCLIB data set includes the text of a couple of books Barry Merrill wrote on SMF data. Chapters 3 and 21 (ACHAP03 and ACHAP21 in the PDS) give an overview of SMF data and CICS workload analysis in particular, respectively. If not, Amazon has Merrill's Expanded Guide to Computer Performance Evaluation Using SAS Systems available hardback for $68.28 according to my system. It is not cheap, but it is the standard for looking at SMF data.