We have recently upgraded from CICS 4.1 to CICS/TS 1.3. In looking around, at nothing in particular, I noticed
a lot of cpu activity with very few tasks running...the bulk of the time was clocked against DFHKETCB. What is that? With Omegamon I am able to look at the load module inside, but it says '*-UNKN-* which I assume means unknown the address is 0000000 so it must be at the beginning of DFHKETCB. Any ideas? I don't know if this is normal or if I have a performance problem.
I don't agree that an idle CICS is going to eat up loads of CPU time if it is not doing anything. If there is nothing to do, everysooften CICS scans the Dispatcher Queues for any work that is outstanding, and if there is none simply goes to sleep. This sleep interval is controlled by the ICV SIT parameter. This is probably what you are seeing against DFHKETCB activity. If CICS has nothing to do, and ICV is set too low, then CICS will be frequently waking up, doing a bit of queue scanning and then going to sleep again. However, the CPU consumed is negligible. See Topics in the Performance Guide about tuning ICV. One thing to bear in mind is that 1.3 does tend to have more asynchronous internal processes occuring than in 4.1. For example, Autoinstalled terminals are logged according to the AI*DELAY SIT parameters and this might be having an effect, but it's unlikely. I cannot comment upon what Omegamon is displaying. I suggest you turn on CICS' Monitoring facilities for these periods of interest and see where the CPU is going. Auxilarily Trace will probably cast some light on things as well. It is possible that some Exits are not properly coded and so are using up extra CPU, but I don't really think this is a potential cause of your observations. There is a well-known low utilisation effect in CICS whereby low activity tends to cost more than in a high activity system, so you might be seeing an effect of this, but I doubt it.
Dig deeper on Mainframe operating systems and management
Related Q&A from Robert Crawford
With 3270 bridge, you can't stack input messages into one structure. The bridge can't process them all at once. It takes a little more work.continue reading
CICS expert Robert Crawford offers advice on determining the connection between CICS transactions and MQ Queue name.continue reading
CICS expert Robert Crawford discusses if a program can use IXLLIST macros in CICS when IXLLIST are APF.continue reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.