Why are highly busy zIIPs worse than highly busy CPs?

[Timothy Sipples]

A potential issue is that some systems only have a single
zIIP.... My personal belief is that given the amount of work
that's zIIP eligible today, most systems should have at least
2 zIIPs....

I would add an important footnote. The current and prior model machine
generations have zIIP engines that are capable of simultaneous
multithreading (SMT). To a first order approximation anyway, one zIIP now
behaves like two when SMT is enabled. Even if you have one zIIP, enabling
SMT2 could be a reasonable way to get the "rule of thumb pair" you
describe.

SMT2 support for zIIPs is available in z/OS 2.1 and higher on IBM z13
machines. Please refer to APARs OA43622 and OA43366.

--------------------------------------------------------------------------------------------------------
Timothy Sipples
IT Architect Executive, Industry Solutions, IBM Z & LinuxONE,
Multi-Geography
E-Mail: ***@sg.ibm.com

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to ***@listserv.ua.edu with the message: INFO IBM-MAIN

[s***@gmail.com]

Going with an SMT will be an good option. Just incase if SMT is enabled in 2.1, It is recommended to go with PTF UA83506 to avoid the system/SRB request to be in a stalled state and the same goes with UA83505 for 2.2

More details at: http://www-01.ibm.com/support/docview.wss?uid=isg1OA51419

.. In addition to SMT, IIPHonorPriority is a good option too. Even if the zIIP is busy, if we have an IIPHONORPRIORITY set to YES, ideally the SRB workloads will be routed to GCP (Setting IIPHONORPRIORTY to NO will park the SRB queues for zIIP processing rather than routing to GCP which will end-up in having a delay or worse situation)