For others that have zEDC card/feature implemented, I am working with a client where we are pursuing "questionable" statistics for SMF30_US_QUEUETIME and SMF30_US_EXECTIME data fields for the SMF type 30 (address space) activity. At this point, IBM is "reviewing the SMF 30 related code." More info to follow when available.
As well, we are working with CA MICS Support to get these metrics added to the MICS SMF, RMF, and VCC/VCA components, so that zEDC-related resources can be tracked/monitored.
Thus far, the MXG support for SMF 30 and 74/9 is accurate, except for fields mentioned above -- however updated MXG SOURCLIB members VMAC30 and VMAC74 are required.
Update to the SMF 30 data source defect: The symptom only occurs with zEDC compression when invoked for SMF LOGSTREAM processing (by LOGR, SMF, and PGM=IFASMFDL address spaces) -- IBM has finally been able to confirm the problem symptom on one of their internal lab z/OS systems. No solution to report but will advise as more information becomes available. On the flip-side, any sequential data copy operations do accurately report SMF 30 zEDC usage section metrics.
Here's an update received late this week - IBM has identified a code problem and the APAR OA48268 is soon to follow for this condition.
IBM has indicated the condition is to be addressed for LOGR/SMF and IFASMFDL (SMF LOGSTREAM OFFLOAD) task executions (started-tasks) as an APAR, in the time-period of 30-Sep-2015 -- I will update as new information is made available.
IBM has now closed this APAR OA45767 to address incorrect SMF 30 zEDC usage section data/statistics reporting specifically for PGM=IFASMFDL (SMF log stream offload processing). Also, there is additional zEDC intermittent failure with S002-F6 ABENDs, which IBM has also corrected in the recent past.
IBM z System sites should at least be aware of SMF log stream exploitation opportunities, benefits, those to include zEDC data compression (all SMF types possibly, not just those supported by say CICS, DB2) -- with SMF data recorded, at one site we support, an 8-to-1 size decrease was observed, as compared to legacy SMF MAN file recording.
Those individuals interested in further discussion can contact me directly.