Now on Monday morning (09:00 PST) page usage for this system looks as 
below. Per SOP, the system was IPLed about 31 hours earlier. I don't know 
when usage began climbing above zero, but I fully expect that by the end 
of the week, it will look similar to my earlier posting. Yes, we run 
health check ASM_LOCAL_SLOT_USAGE. And yes, it begins to complain when any 
local grows above 30%. 

Note that this paging configuration is quite static. It hasn't changed for 
many months. Certainly last week (previous post) there were no PAGE ADDs 
or DELETEs. What I do notice now (missed completely before!) is that for 
each even/odd numbered pair, both on the same unit address on the same 
DS8* subsystem, usage of the first (even) data set is higher than the odd 
data set. In *all* cases I created each pair at more or less the same time 
using the same IDCAMS job. This usage skew, seen here with minor 
differences, increases through the week (previous post). BTW every IPL is 
CLPA. 

Some time last year I added several GB of memory to this LPAR. That 
reduced paging activity, e.g. fewer alerts from Omegamon, but did not have 
much effect on page data set usage. 

LOCAL     18%   OK  1208  SYS1.PAGELOC0
LOCAL     10%   OK  1208  SYS1.PAGELOC1
LOCAL     16%   OK  1308  SYS1.PAGELOC2
LOCAL     10%   OK  1308  SYS1.PAGELOC3
LOCAL     18%   OK  1207  SYS1.PAGELOC4
LOCAL     11%   OK  1207  SYS1.PAGELOC5
LOCAL     17%   OK  1218  SYS1.PAGELOC6
LOCAL     10%   OK  1218  SYS1.PAGELOC7

.
.
JO.Skip Robinson
Southern California Edison Company
Electric Dragon Team Paddler 
SHARE MVS Program Co-Manager
626-302-7535 Office
323-715-0595 Mobile
[email protected]



From:   "[email protected]" <[email protected]>
To:     [email protected], 
Date:   03/03/2013 10:27 PM
Subject:        Paging Behaviour; was: Re: DFSORT Weirdness
Sent by:        IBM Mainframe Discussion List <[email protected]>



> All data sets are the same size at 5,008 cylinders, half of a 3390-9. 
> 
> LOCAL     55%   OK  1208  SYS1.PAGELOC0
> LOCAL     34%   OK  1208  SYS1.PAGELOC1
> LOCAL     51%   OK  1308  SYS1.PAGELOC2
> LOCAL     32%   OK  1308  SYS1.PAGELOC3
> LOCAL     53%   OK  1207  SYS1.PAGELOC4
> LOCAL     37%   OK  1207  SYS1.PAGELOC5
> LOCAL     53%   OK  1218  SYS1.PAGELOC6
> LOCAL     35%   OK  1218  SYS1.PAGELOC7

> I can remember (but not pinpoint) when usage stayed 
> very low all week long.

Two things come to my mind on this: 

I have only ever seen wildly different numbers like yours in page data set 
percentage usage (when they are all the same size) when page data sets 
were added later, when the rest already had a certain usage. It seemed 
that adding a page ds doesn't favour that page ds so it reaches the same 
threshold as the others, but instead it gets pages in round-robin fashion, 
slowly building up percentage usage, but never reaching the usage of the 
previous page ds's. In our case, all page ds were the same size, too, and 
were all behind the same controller. It took an IPL to even out usage.

Do you have that CHECK(IBMASM,ASM_LOCAL_SLOT_USAGE) health check active on 
that system? I once used the exception thrown via the hardcopy log 
messages to pinpoint that usage increased way before I thought it had 
increased. It was a gradual creep.

I had also seen similar (in my opinion strange) behaviour, and in essence 
I established a reporting network for storage usage, including total 
amount of slots and paging rates. By now, that is quite a trend grafic. 
Some of the behaviour visible in those grafics was also quite unexpected. 
Mostly, I used the numbers to get those lpars a bit more real storage when 
we migrated to 1.12 and had several RSM problems that never were found. 
They went away with just a bit more real storage. But I had to fight tooth 
and nail for it. One would think that that real storage was intended for 
my personal enjoyment, the way my management kept withholding it for z/OS 
while throwing it at z/Linux....

Barbara


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

Reply via email to