Response time needs to be measured at 2 locations for proper problem
determination. These 2 places are the boundary of the CPU and the
NETWORK and the end-point of the communication.

You indicated that CPU consumption by the JES address space, attributed
to RJE processing seems to be causing the delay.

If the incremental CPU consumed by JES is enough to make a difference,
this implies a really enormous CICS workload. At any rate, 
as far as I know, there is now way to identify a "RJE TCB" within the
JES address space to workload manager, so this is a non-starter. 
<editorial>IMO, applying any limit to JES is a *VERY BAD*
idea.</editorial>

If the local response (measured by CICS transaction statistics) is
acceptable, this would indicate a communications delay. A communications
link shared between "online" and batch) seems a more likely cause than
incremental CPU consumption. The severity of the problem would be
inversely proportional to the speed of communication link.

There are 2 solutions that come to mind for a "saturated" communications
link. The first and logically the easiest, is to put the RJE traffic on
a different link, however this may not be financially feasible. The
second is a "Class of Service" implementation (SNA or TCP, you are out
of luck if BiSYNC). There are extensive description of a  "Class of
Service" implementation in the communications server manuals

http://publibz.boulder.ibm.com/cgi-bin/bookmgr_OS390/Shelves/F1A1BKB0

HTH,

<snip>
Recently, we have a problem when sending large amount spool files (about
2.5
million lines) to RJE printer, as the work caused JES2 consumes lots of
CPU,
therefore impacted
the CICS online performance.
We consider to limit JES2 address space CPU, using WLM resource group,
but
this will suppress other JES2 process as well. So, any idea to limit RJE
work only?
</snip>

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to