Chapter 29 Exercises

Section 29.2 — Measuring Capacity

Exercise 29.1: Unit Conversion

A z16 Model A01-706 has a rated capacity of 6,534 MSU. Using the approximate conversion of 8.5 MIPS per MSU:

a) Calculate the approximate MIPS rating of this machine. b) If a COBOL batch program consumes 12,500 CPU seconds during a nightly run on this machine, what percentage of one hour's total GP capacity does that represent? (Assume the program runs within a single hour.) c) A vendor claims their z/OS software product "only uses 50 MIPS." Convert this to MSU and calculate the approximate annual MLC cost at $15/MSU/month. d) Explain why a CPU second on a z14 and a CPU second on a z16 represent different amounts of "work," even though both are one second of processor time.

Exercise 29.2: RMF Duration Report Analysis

Given the following RMF Post-Processor duration report for a four-LPAR sysplex:

LPAR    Engine  Avg%   Peak%   Peak Time           MSU(Avg)  MSU(Peak)
------  ------  -----  ------  ------------------  --------  ---------
PRDA    GP      71.4   93.2    2025-03-31 02:15    1,245     1,625
PRDA    zIIP    52.3   81.7    2025-03-31 02:30    n/a       n/a
PRDB    GP      63.8   87.1    2025-03-15 01:45    1,113     1,520
PRDB    zIIP    44.1   72.6    2025-03-15 02:00    n/a       n/a
PRDC    GP      48.2   71.3    2025-03-31 13:30    841       1,244
PRDC    zIIP    31.5   55.8    2025-03-31 14:00    n/a       n/a
PRDD    GP      39.7   62.4    2025-03-15 13:00    693       1,088
PRDD    zIIP    22.8   41.2    2025-03-15 13:30    n/a       n/a

a) Identify the LPAR with the highest peak utilization. Is this during the batch window or the online window? How do you know? b) Calculate the average GP utilization across all four LPARs. Is the workload balanced? c) What is the total sysplex average MSU consumption? What is the peak? d) PRDA hits 93.2% peak GP utilization. Explain why this is a capacity concern, referencing WLM's behavior above 85% utilization. e) Propose a workload rebalancing strategy that would reduce PRDA's peak utilization below 85% without increasing any other LPAR above 80%. f) The peak times suggest two distinct peak periods — overnight (01:00-03:00) and midday (13:00-14:30). What does this tell you about the workload mix?

Exercise 29.3: zIIP Offload Analysis

A shop runs the following workloads on their z16:

a) Calculate the total daily GP CPU seconds being consumed. b) Calculate the total daily CPU seconds that could be running on zIIP but are currently running on GP (the offload gap). c) If the offload gap were fully closed (all eligible work moved to zIIP), what would the daily GP CPU reduction be? d) At an approximate conversion of 54.5 CPU seconds per MSU per hour, estimate the peak MSU reduction from full zIIP offload (assume the peak hour represents 1/8 of daily volume). e) At $15/MSU/month MLC cost, what is the approximate annual savings from full zIIP offload? f) Which single workload offers the largest absolute MSU savings opportunity? Which offers the largest percentage improvement?

Exercise 29.4: SMF Data Collection Design

You are designing the SMF data collection strategy for a new capacity planning repository. The shop runs 3 LPARs with the following characteristics:

• 1,800 batch jobs per night (average 3.5 steps each)
• 45,000 online transactions per hour during peak (8 hours of peak)
• RMF Monitor I interval: 15 minutes

a) Calculate the number of SMF type 70 records generated per day (one per RMF interval per LPAR). b) Calculate the number of SMF type 72 records generated per day, assuming 12 WLM service classes per LPAR. c) If each type 70 record averages 2,400 bytes and each type 72 record averages 1,800 bytes, what is the daily SMF volume for capacity planning records alone? d) You want to retain hourly data for 90 days and daily aggregates for 3 years. Design the DB2 table structure (table names, key columns, and primary metrics for each table). e) Estimate the DB2 storage required for 3 years of data at these volumes.

Exercise 29.5: Capacity Data Pipeline

Write the COBOL working storage definitions for a program that reads SMF type 70 (Processor Activity) records and extracts the following for each LPAR:

• System ID
• LPAR name
• Measurement interval start time
• Measurement interval end time
• GP busy percentage
• zIIP busy percentage
• GP MSU consumption
• Total LPAR weight and capped status

Include the necessary record layout for SMF type 70 header fields and the processor section. You do not need to write the full program — just the file and working storage sections.

Section 29.3 — Workload Characterization

Exercise 29.6: Workload Profile Construction

Given the following WLM service class data for one LPAR (monthly summary):

Service Class     Avg CPU/sec/hr   Peak CPU/sec/hr   Peak Time
────────────────  ──────────────   ───────────────   ─────────
SC_CICS_HIGH      8,400            14,200            12:30 PM
SC_CICS_NORMAL    3,200            5,800             12:15 PM
SC_CICS_LOW       1,100            2,400             11:45 AM
SC_BATCH_CRIT     2,800            18,500            01:30 AM
SC_BATCH_STD      1,500            9,200             02:00 AM
SC_BATCH_LOW      400              3,100             03:15 AM
SC_DB2_UTIL       800              6,400             03:30 AM
SC_DB2_DRDA       2,100            4,800             12:45 PM
SC_MQ_HIGH        600              1,200             01:00 AM
SC_MQ_STD         300              800               12:00 PM
SC_TSO            900              1,800             10:30 AM

a) Calculate the total average CPU consumption per hour across all service classes. b) Calculate the daytime peak (using online, DB2 DRDA, MQ, and TSO service classes). c) Calculate the nighttime peak (using batch, DB2 utility, and MQ service classes). d) What percentage of total average CPU is consumed by batch processing? By online processing? e) If online workloads grow at 15% per year and batch grows at 8%, what will the daytime and nighttime peaks be in 18 months? f) At what point (in months from now) will the daytime peak exceed the current nighttime peak? What are the implications for capacity planning?

Exercise 29.7: Workload Mix Shift Analysis

A shop has tracked its workload mix over five years:

a) Calculate the year-over-year change in each category's share. b) If these trends continue, project the workload mix for 2025 and 2026. c) At what year does API/Integration processing become the largest workload category? d) Explain why this mix shift matters for MSU budgeting, considering that API/Integration workloads are significantly more zIIP-eligible than batch workloads. e) If the shop currently has 4 zIIP engines and 12 GP engines, at what point should they request additional zIIPs? What ratio of zIIP to GP would you recommend for 2026?

Exercise 29.8: Cross-LPAR Workload Analysis

A four-LPAR sysplex runs the following workloads:

LPAR    Primary Workload          GP MSU (Avg)  GP MSU (Peak)  Peak Time
────    ────────────────          ────────────  ─────────────  ─────────
PRDA    CICS Production + Batch   1,245         1,625          02:15 AM
PRDB    CICS Production + Batch   1,113         1,520          01:45 AM
PRDC    DB2 + API Layer           841           1,244          01:30 PM
PRDD    Development/Test          693           1,088          01:00 PM

a) Calculate the sysplex peak MSU. Is this the sum of all LPAR peaks? Why or why not? b) If all four LPARs peak at the same time (worst case), what would the sysplex peak be? c) Propose a strategy to reduce PRDA's peak by redistributing batch workload to PRDB or PRDD. d) PRDD runs development/test with a peak of 1,088 MSU at 1:00 PM. Propose a WLM capping strategy that limits PRDD to 800 MSU during the nighttime batch window (11 PM - 6 AM) to free capacity for PRDA/PRDB. e) Calculate the R4HA impact of the capping strategy from (d), assuming the freed capacity reduces PRDA's peak by the amount capped from PRDD.

Section 29.4 — Forecasting Models

Exercise 29.9: Linear Regression Forecast

You have 24 months of R4HA data (monthly):

Month  R4HA(MSU)  Month  R4HA(MSU)  Month  R4HA(MSU)  Month  R4HA(MSU)
  1     2,810       7     2,930      13     3,050      19     3,180
  2     2,780       8     2,900      14     3,010      20     3,140
  3     2,850       9     2,960      15     3,080      21     3,220
  4     2,830      10     2,990      16     3,060      22     3,200
  5     2,870      11     3,020      17     3,100      23     3,240
  6     2,920      12     3,100      18     3,160      24     3,310

a) Calculate the linear regression slope (MSU per month) using the least-squares method, or estimate it from the data by comparing month 1 to month 24. b) Project the R4HA for months 30, 36, and 48. c) At what month does the projected R4HA exceed 4,000 MSU? At what month does it exceed the machine's rated capacity of 4,500 MSU? d) If a hardware upgrade takes 6 months from decision to installation, when must the upgrade decision be made to avoid exceeding 4,000 MSU? e) What factors make this linear projection unreliable beyond 18 months?

Exercise 29.10: Seasonal Index Calculation

Using the 24 months of data from Exercise 29.9:

a) Calculate the 12-month moving average for months 7–18 (the months with a full year of data before and after). b) Calculate the seasonal index for each month by dividing each data point by the 12-month average centered on that month. c) Average the seasonal indices for each calendar month (January = months 1 and 13, February = months 2 and 14, etc.). d) Which month has the highest seasonal index? Which has the lowest? e) Using your seasonal indices and the linear trend from Exercise 29.9, forecast the R4HA for each month of Year 3 (months 25–36). f) What is the difference between the highest and lowest monthly forecast in Year 3? Express this as a percentage of the annual average.

Exercise 29.11: Business Event Capacity Impact

Your shop is planning the following business events for the next 18 months:

Current baseline: 3,310 MSU (GP), 800 MSU (zIIP).

a) Build a composite forecast that combines the linear trend from Exercise 29.9, the seasonal indices from Exercise 29.10, and these business events. b) Produce three scenario forecasts (conservative, expected, aggressive) for months 25–36. Define your assumptions for each scenario. c) At what month does the conservative scenario first exceed 4,000 MSU? The expected? The aggressive? d) For the cloud migration (Month 15, Low confidence): explain why low confidence increases the planning range and describe how you would improve confidence. e) Produce the MSU budget for months 25–36, including CoD requirements for months where the expected scenario exceeds current capacity.

Exercise 29.12: Forecast Accuracy Assessment

After 12 months, you compare your forecasts to actuals:

Month   Forecast  Actual   Variance   Variance %
  1      3,350    3,310    -40        -1.2%
  2      3,320    3,280    -40        -1.2%
  3      3,470    3,510    +40        +1.2%
  4      3,400    3,440    +40        +1.2%
  5      3,430    3,380    -50        -1.5%
  6      3,820    4,020    +200       +5.2%
  7      3,510    3,530    +20        +0.6%
  8      3,480    3,500    +20        +0.6%
  9      3,490    3,320    -170       -4.9%
 10      3,560    3,610    +50        +1.4%
 11      3,600    3,630    +30        +0.8%
 12      3,730    3,780    +50        +1.3%

a) Calculate the Mean Absolute Percentage Error (MAPE) for the full 12 months. b) Month 6 shows a +5.2% variance. This was the acquisition integration month. What does this tell you about your capacity impact estimate for that event? c) Month 9 shows a -4.9% variance. This was the DB2 upgrade month. The forecast assumed a 5% CPU reduction. What actually happened? d) Excluding the two event months (6 and 9), recalculate the MAPE. How does this change your assessment of the baseline forecasting model? e) What adjustments would you make to the forecasting model for the next year based on this accuracy assessment?

Section 29.5 — MSU Budgeting and Optimization

Exercise 29.13: R4HA Calculation

Given the following hourly MSU consumption for a single LPAR over a 24-hour period:

Hour   MSU    Hour   MSU    Hour   MSU    Hour   MSU
00:00  2,100  06:00  1,800  12:00  2,400  18:00  2,000
01:00  2,800  07:00  2,100  13:00  2,500  19:00  1,900
02:00  3,200  08:00  2,200  14:00  2,300  20:00  1,800
03:00  3,400  09:00  2,300  15:00  2,200  21:00  1,700
04:00  3,100  10:00  2,400  16:00  2,100  22:00  1,900
05:00  2,600  11:00  2,350  17:00  2,100  23:00  2,000

a) Calculate the rolling four-hour average for every possible four-hour window. b) Identify the R4HA (highest four-hour average). At what time does the window start? c) If you could shift 300 MSU of batch workload from the 02:00-04:00 period to the 22:00-00:00 period, recalculate the R4HA. What is the savings? d) Calculate the annual MLC savings from this R4HA reduction at $15/MSU/month. e) Is there a further scheduling optimization that could reduce the R4HA below 2,800 MSU? Describe it and calculate the result.

Exercise 29.14: TFP vs. MLC Decision

A shop is evaluating IBM's Tailored Fit Pricing (ECS model) versus traditional MLC. Their current profile:

• Average R4HA: 2,800 MSU
• Peak R4HA (month-end): 3,400 MSU
• Number of peak months per year: 4 (quarter-ends)
• Current MLC rate: $14/MSU/month
• TFP/ECS commitment: 3,000 MSU at $11/MSU/month
• TFP/ECS overage rate: $18/MSU/month for consumption above commitment

a) Calculate the annual MLC cost under the current model (using the peak R4HA × 12 months, since MLC uses the monthly peak). b) Wait — that's not correct. MLC uses the monthly peak R4HA, which varies. Calculate the annual cost more accurately: - 4 months at 3,400 MSU - 8 months at 2,800 MSU c) Calculate the annual TFP/ECS cost: - 12 months of base commitment (3,000 MSU × $11) - 4 months of overage (400 MSU × $18) - 8 months where consumption is below commitment (no refund) d) Which model is cheaper? By how much annually? e) Under what circumstances would the other model become cheaper? (What R4HA profile would tip the balance?)

Exercise 29.15: zIIP Optimization Business Case

You are proposing a z/OS Connect implementation that will shift balance inquiry transactions from CICS local DB2 calls to DRDA-based API calls. The capacity impact:

Current state: - 15M balance inquiries/day - 0.0030 GP CPU seconds/transaction - 0% zIIP eligible

Proposed state: - 15M balance inquiries/day via z/OS Connect - 0.0038 total CPU seconds/transaction (higher due to API overhead) - 72% zIIP eligible

a) Calculate the current daily GP CPU consumption. b) Calculate the proposed daily GP CPU consumption (only the non-zIIP portion). c) Calculate the proposed daily zIIP CPU consumption. d) Calculate the GP MSU reduction (approximate, using 54.5 CPU sec/MSU/hour and assuming even distribution across 12 peak hours). e) At $15/MSU/month, calculate the annual MLC savings. f) The z/OS Connect implementation costs $250,000 (one-time) plus $80,000/year in licensing. Calculate the ROI and payback period. g) What risks should you include in the business case? List at least three.

Section 29.6 — Capacity Impact Analysis

Exercise 29.16: New Application Impact Assessment

A new fraud detection module is being deployed to the CICS production environment. Testing shows:

• Additional CPU per transaction: 0.0008 seconds
• 40% of transactions will invoke the fraud check: 200M transactions/day × 40% = 80M
• 30% of fraud check CPU is zIIP-eligible (ML model inference via Java)
• The module adds 2 DB2 SELECTs per invocation (average 15 getpages each)

a) Calculate the daily GP CPU increase. b) Calculate the daily zIIP CPU increase. c) Estimate the peak MSU impact (assume the peak hour has 12% of daily volume). d) What is the impact on the R4HA if the fraud module runs during the online peak (12:00-4:00 PM)? e) Complete a Capacity Impact Assessment using the template from Section 29.6.4.

Exercise 29.17: DB2 Index Impact Analysis

Lisa Tran is evaluating a request to add a composite index on the TRANSACTION_HISTORY table:

Table:          TRANSACTION_HISTORY
Current size:   2.4 billion rows
Daily inserts:  12 million rows
Daily deletes:  0 (archival is weekly)
Row length:     380 bytes
Current indexes: 2 (primary key, date-based)

Proposed index: (ACCOUNT_ID, TX_TYPE, TX_DATE DESC)
Index entry:    48 bytes per row

Queries that would benefit: - TXNINQUIRY: 50,000 executions/day, currently 800 getpages/exec, with index: 12 getpages/exec - TXNREPORT: 200 executions/day, currently 45,000 getpages/exec, with index: 2,500 getpages/exec

a) Calculate the current daily CPU for the two queries (estimate 0.0001 CPU seconds per getpage for sequential access, 0.00005 per getpage for index access). b) Calculate the projected daily CPU with the new index. c) Calculate the additional CPU for index maintenance on 12M daily inserts (estimate 0.0002 CPU seconds per insert per additional index). d) What is the net CPU impact (positive or negative)? e) Calculate the DASD storage required for the new index (2.4 billion entries × 48 bytes, with 20% free space for inserts). f) Should Lisa approve this index? Justify your answer with the capacity impact data.

Exercise 29.18: Platform Migration Impact

Sandra Chen at Federal Benefits is moving the benefits inquiry workload from IMS to a cloud-hosted microservice via z/OS Connect. Current and projected profiles:

a) Calculate the current daily GP CPU for IMS-based inquiry. b) Calculate the projected daily GP CPU for the z/OS Connect bridge model. c) Calculate the projected daily zIIP CPU. d) Calculate the net GP MSU reduction. e) If the MLC savings from GP reduction are $X/year and the cloud service costs $180,000/year, what minimum annual MLC savings makes the project financially viable? f) What non-financial factors should Sandra consider in this capacity impact assessment?

Section 29.7 — The Capacity Planning Process

Exercise 29.19: Annual Capacity Plan Construction

You are the capacity planner for a mid-size insurance company. Construct the outline of an annual capacity plan given:

• Current capacity: 2,200 MSU (GP), 450 MSU (zIIP)
• Machine: z15 T02 rated at 3,300 MSU
• Current utilization: 67% average, 82% peak
• Historical growth: 10% per year for 3 years
• Business events next year:
• New claims portal (Q2): estimated +120 MSU online
• Regulatory reporting expansion (Q3): estimated +80 MSU batch
• Analytics platform (Q4): estimated +200 MSU zIIP
• Seasonal pattern: Q4 is 15% above average (open enrollment)

a) Produce the three-scenario forecast (conservative, expected, aggressive) by quarter for the next fiscal year. b) At what point does the expected scenario exceed 85% of the machine's rated capacity? c) At what point does the aggressive scenario exceed the machine's rated capacity? d) Propose a procurement timeline that ensures capacity is available before the 85% threshold is breached. e) Calculate the MSU budget with quarterly projections. f) Define the exception thresholds for your capacity monitoring. g) Draft the executive summary (2-3 paragraphs) that you would present to IT leadership.

Exercise 29.20: Quarterly Variance Analysis

It's the end of Q1 and you're comparing actuals to your plan:

                 Plan (Expected)   Actual    Variance
January R4HA:    2,320             2,280     -1.7%
February R4HA:   2,300             2,410     +4.8%
March R4HA:      2,380             2,520     +5.9%
GP Avg Util:     68%               71%       +3 pp
zIIP Avg Util:   45%               52%       +7 pp
Batch Window:    82% used          87% used  +5 pp

a) Which metrics indicate a potential capacity concern? b) The February and March R4HA exceeded the Expected scenario. What are possible explanations? c) zIIP utilization increased faster than planned (+7 pp). Is this good news, bad news, or neutral? Explain. d) The batch window is using 87% of available time vs. 82% planned. Connect this to the capacity analysis — what's consuming the additional 5 percentage points? e) Draft the recommended actions section of your quarterly capacity review.

Exercise 29.21: Capacity Governance Design

Design a capacity governance framework for a shop that currently has no formal capacity planning process. Include:

a) Roles and responsibilities (who does what). b) The annual planning calendar (month by month). c) Data collection requirements (what SMF/RMF data, retention periods). d) Review cadence (annual, quarterly, monthly, exception-driven). e) The capacity impact assessment process (when is a CIA required, who approves, what's the turnaround time). f) Integration points with other processes (change management, project planning, financial planning, procurement). g) Success metrics (how do you know the capacity planning process is working).

Exercise 29.22: Capacity Exception Response

At 2:47 AM on a Tuesday, your capacity monitoring alerts fire:

• PRDA GP utilization has been above 92% for the past 45 minutes
• The R4HA for the current month just exceeded the Aggressive scenario threshold
• Batch window progress is 12 minutes behind schedule

a) Classify each alert using the three-tier exception framework from Section 29.7.3. b) What immediate actions should the on-call capacity analyst take? c) What data would you collect during the incident for post-incident analysis? d) After the incident resolves, you discover that a new batch job (deployed yesterday by the application team without a CIA) is consuming 400 CPU seconds per execution and runs 150 times. Calculate the CPU impact and explain why this wasn't caught. e) Draft a capacity governance improvement recommendation based on this incident.

Exercise 29.23: Multi-Year Capacity Roadmap

Using the following inputs, build a three-year capacity roadmap:

Current state (Year 0): - z15 T02, 3,300 MSU rated capacity - Current consumption: 2,200 MSU GP (R4HA), 450 MSU zIIP - Growth rate: 10% GP, 20% zIIP

Business pipeline: - Year 1 Q2: Cloud migration phase 1 (-150 MSU GP, +80 MSU zIIP) - Year 1 Q4: Acquisition (+400 MSU GP, +100 MSU zIIP) - Year 2 Q2: API modernization (-200 MSU GP batch, +300 MSU zIIP API) - Year 2 Q4: AI/ML fraud detection (+50 MSU GP, +500 MSU zIIP) - Year 3 Q2: Cloud migration phase 2 (-300 MSU GP, +120 MSU zIIP)

a) Plot the GP and zIIP consumption quarterly for all three years. b) At what quarter does GP consumption exceed the z15's rated capacity? c) At what quarter does zIIP consumption become a concern (assume 6 zIIP engines with 550 MSU capacity each = 3,300 MSU total)? d) Recommend a hardware upgrade path (when to upgrade, what to upgrade to). e) Calculate the total cost of ownership for the three-year period, including hardware depreciation ($1.2M/year for z15, estimated $1.8M/year for z16), MLC software ($14/MSU/month), and CoD charges for interim capacity. f) Present an alternative scenario where aggressive zIIP offload defers the GP hardware upgrade by one year. What offload targets are required?

Progressive Project

Exercise 29.24: HA Banking System Capacity Plan Review

Review the capacity plan created in Section 29.8 of this chapter. Identify and address:

a) Three assumptions that you would challenge or want to validate with additional data. b) Two additional business events that should be on the capacity events calendar for a banking transaction processing system. c) The CoD strategy for bridging the gap between Y1-Q4 (when capacity gets tight) and the hardware upgrade completion. How many CoD days, at what MSU level? d) The DR capacity requirements — if the HA Banking System fails over to the DR site (Chapter 30), what MSU must be available at the DR site? Is this the full production capacity or a subset? e) The monitoring integration — what capacity metrics from this plan should feed into the batch monitoring system from Chapter 27?

Exercise 29.25: Capacity Optimization Proposal

Using the HA Banking System workload profile from Section 29.8, propose a capacity optimization plan that defers the hardware upgrade by 12 months. Your plan must include:

a) zIIP offload targets (which workloads, how much offload, what technology changes required). b) R4HA management actions (batch scheduling changes, WLM capping, workload balancing). c) Application-level optimizations (DB2 tuning, COBOL compiler upgrade, batch consolidation). d) Financial analysis (cost of optimizations vs. cost of earlier hardware upgrade vs. savings from deferral). e) Risk analysis (what could go wrong with the deferral strategy, and what triggers should cause you to accelerate the upgrade). f) A decision matrix that quantifies the deferral option against the on-schedule upgrade option.