Colin Allinson wrote:
The background to this question is that we have a very busy IBM
2084/B16-307 running 6 active VM partitions - but one of them, VM2, is
much more heavily used than the others.
We used to share all 7 engines across all partitions (uncapped) and
leave it to the partition weighting to sort out the relative
priorities. A few weeks ago we were asked to reduce the number of
engines used by the minor partitions in order to favour the key
systems, in particular VM2.
Yesterday we had performance problems on VM2 and what I saw surprised
me :-
1. All 7 engines on the processor were running at (or near)
100% (no surprise here).
2. The HMC SAD showed that VM2 was using 129% of its allocated
share (no surprise here either).
3. CP IND on VM2 was showing it was running at 99% (100% on
PROC 0). Clearly this was the cause of the performance issue !!
4. ESAMON and PERF Toolkit showed that VM2 was using just under
64% of each of the engines.
What surprises me is that VM2 is showing 100% utilisation when it was
uncapped and only getting 64% of the total resources. I am 99% sure
this is not what we were seeing before we changed the allocation of
the processors. It looks to me as if we have somehow unwittingly
introduced a capping on VM2 of 64% of the total processor resources.
The partition controls are as follows:-
Partition Active Weight Min Weight Max Weight Capping
Non-Dedicated CPUs
VML Yes 250 0
0 No 7
VMT Yes 150 0
0 No 3
VMX Yes 70 0
0 No 3
VM1 Yes 70 0
0 No 4
VM2 Yes 750 0
0 No 7
VM3 Yes 150 0
0 No 7
VM5 No 40 0
0 No 2
Can anyone throw any light on this?
Thanks in advance
Colin Allinson
Amadeus Data Processing
Colin
This sounds like a Master processor bottleneck. What is your z/VM level?
Phil
--
'in media stat virtus'
Virtue's in the middle
