<snip>
Enclaves were supposed to be an exception to this rule. First, SRB's in an 
enclave are interruptible. Second, for dispatching purposes, SRB's and 
TCB's placed in an enclave are excluded from the address space. The system 
will select either an enclave or address space for processing.  Time 
slicing applies to the selected enclave or address space (excluding 
enclave SRB's and TCB's). Essentially a single address space acting like 
multiple address spaces. A real solution for a mixed workload address 
space because CHAP and WLM do not solve the problem.

Can you tell us the reality versus what I was told about enclaves?  For 
the the product I worked on, we discussed implementing enclaves but the 
workload could not be classified correctly. I only have research 
experience with WLM enclaves.
</snip>

This is the 2nd time in a row that you were responding to only part of 
what was written, whether intentionally or not. Yes, enclaves are an 
exception. My "rule" used the term "equal", within "all other things being 
equal". Thus it would not include cases that were unequal, whether that be 
CHAP or enclaves or something else. 

Regardless, your understanding seems flawed.

First, time slicing does not change based on "selected enclave or address 
space".  There may be different time slices for zIIP work vs CP work, or 
when HiperDispatch=YES is in effect vs HiperDispatch=NO. But the time 
slice does not vary based on the enclave or the address space.

Second, everything comes down to priority. WLM does its thing by managing 
the priorities. And dispatch is done based on the resulting priority.

There is a "major priority" (0-255) and a "minor priority" (0-255). Think 
of the major as the address space priority. Think of the minor priority as 
the task priority. Aside from edge cases, things are dispatched in 
priority order. The minor priority comes into play only when comparing 
things with equal major priority.

That leads to the question: where does "enclave priority" fit in? As part 
of "major".  The major priority for a work unit comes from the owning 
enclave (if the work unit is in an enclave) or the owning address space 
(if the work unit is not in an enclave).

When a work unit is undispatched while still ready (I think also for the 
case of "made ready") it is placed at the end of things of equal priority 
(that's how the approximate round-robin is accomplished).

This is not the "complete rule". There are tweaks and exceptions in 
various places.

Peter Relson
z/OS Core Technology Design


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

Reply via email to