Specialty Engines and
Kneecapped Processors
The Myth of MSU II
Jim Horne
What We Will Cover
• Definitions
• What we can measure
– TYPE 70 information
– TYPE72 information
• What the information tells us
• What we can learn from it
Definitions
• Specialty Engines
– zIIPs, zAAPs
– Ignore other types as not applicable to z/OS
• GPs
– general purpose processors
• Kneecapped processors
– Processors whose GPs do not run at full speed
What We Can Measure
TYPE 70
• SMF70NRM
• SMF70CIN
• SMF70CTN
•
•
•
•
SMF70CAN
SMF70DSA
SMF70IFA
SMF70SUP
normalization factor for zIIP
CPU type (CP, IFL, AAP, IIP, ICF)
CPs of type SMF70CIN online at
End of Interval (EOI)
accumulated CPs during interval
number of diagnose samples
number of online zAAPs at EOI
number of online zIIPs at EOI
What We Can Measure
TYPE 70
•
•
•
•
•
SMF70PAT
CPU parked time
SMF70WAT CPU wait time
SMF70PDT
LP dispatch time
SMF70EDT
LP effective dispatch time
SMF70CIN (again)
What We Can Measure
TYPE72
• R723MADJ
•
•
•
•
R723MCPU
R723MSRB
R723NFFI
R723NFFS
Adjustment factor for CPU rate
(service units per second)
CPU service coefficient * 10,000
SRB service coefficient * 10,000
zAAP normalization factor
zIIP normalization factor
What We Can Measure
TYPE72
•
•
•
•
R723CSUC
R723CIFC
R723CSUP
R723CIFA
SUs on GPs for zIIP eligible work
SUs on GPs for zAAP eligible work
SUs on zIIPs
SUs on zAAPs
What the information tells us
• TYPE70
– Actual engine use
– Nothing about eligible specialty engine use
– Kneecapping factor
• TYPE72
– Service class engine use
– Actual and eligible specialty engine use
– Kneecapping factor
What the information tells us
• Kneecapping factor
– xxxxxxxx / 256
– TYPE70 – SMF70NRM
– TYPE72
• R723NFFI (zAAP)
• R723NFFS (zIIP)
• Service units to seconds
– R723MADJ
Table 1 - Selected Processor Ratios
Processor
2097-4xx
2097-5xx
2097-6xx
2097-7xx
2817-4xx
2817-5xx
2817-6xx
2817-7xx
Full Size
Ratio
4.21
1.95
1.44
1.00
4.89
2.00
1.53
1.00
What we can learn from them
One CPU sec on GPs v. Specialty
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
2097-4xx 2097-5xx 2097-6xx 2097-7xx 2817-4xx 2817-5xx 2817-6xx 2817-7xx
GP Time
Specialty Time
One CPU sec on Specialty v. GPs
5.00
4.50
4.00
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0.00
2097-4xx 2097-5xx 2097-6xx 2097-7xx 2817-4xx 2817-5xx 2817-6xx 2817-7xx
GP Time
Specialty Time
The Myth of MSU II
Presentation does not matter
Engines
• Management buys engines
• Present data to them in terms of engines
• How to calculate:
(CPU time) / Duration = Engines
1 SP + 1 GP on a 2097-606
2.50
2.00
1.50
GP Use
Specialty Use
1.00
0.50
0.00
Observed
As if all on GP
As if all on Specialty
1 SP + 1 GP on a 2817-402
6.00
5.00
4.00
GP Use
3.00
Specialty Use
2.00
1.00
0.00
Observed
As if all on GP
As if all on Specialty
Final Thoughts
• We need to understand kneecapping and its
effect on reporting specialty engine use
• We need to understand our data and what it is
telling us
• We need to show what is actually happening
• We need to be able to discuss “What if…?”
– How many GPs we might save
– How many specialty engines we might need