I created an SHARE requirement (if you're a member you can find it at this 
link: https://www.share.org/p/is/vi/iid=154&type=12 ) about this, but I'd love 
to hear some discussion so I'm posting here too.  My idea is that when IBM 
offers some new "outboard" capability such as zEDC, they need to provide SPE 
PTFs to add measurement of the "old way" so we can use data to decide whether 
to order, and of course pay for, the new feature.  For example, in the zEDC 
case it would have been nice to have a measurement of the CPU time used for SMS 
compression.   If you haven't voted on the requirement, please do so, but 
whether you do or you don't I'd like to hear opinions and ideas.

A second part of the "outboard" measurement issue occurs to me and it isn't 
_just_ hardware related.   While we have a huge amount of information in the 
SMF type 30 records,  CPU use is still scattered among many records.  For 
example, anyone who uses DB2 at their shop knows that the DB2 CPU time is part 
of the CPU time recorded in the SMF type 30 - but how _much_ of the CPU time is 
due to DB2? The record doesn't tell us.   I think it's time for IBM to think 
about some sort of CPU measurement recording "architecture"  - maybe a section 
of type 30, maybe a new SMF record - that gathers in one place the CPU 
measurements of all 'subsystems' used by a job/transaction/enclave  where 
'subsystem' can be a software piece _or_ a hardware piece. Record the CPU time 
for use GP, zIIP, crypto engine, zEDC engine, IDAA card,  in-DB2 time, time in 
CICS code, time in IMS DL/1 processing, whatever.  I'm no expert and I know 
there are going to be nuances (what if my job uses zEDC _and_ my in-DB2 time 
uses zEDC? for one thought) - but we have a lot of nuances now, too.  Again, 
opinions and ideas are what I'm looking for.

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to