1. zSeries, Sub-Capacity Workload License Charges, Soft-Caps, and WLM
Richard S. Ralston
Humana, Inc.

2. Contents What is Sub-Capacity License Charges Tools
Observations and Behaviors
Managing the LPAR Defined Capacity
Benefits
Conclusions/Results
References
Contact Info

3. What is Sub-Capacity License Charges
Software cost charging methodology based upon monthly usage and LPAR size definitions (LPAR Defined Capacity)
z/Series and z/OS only
VWLC software only (z/OS, DB2, CICS, Netview, COBOL, RACF, etc.)

4. How Sub-Capacity License Charges Work
Define the LPAR capacity in MSUs
Usage is tracked in a 4 hour rolling average
Charges are based upon the highest 4 hour rolling average or the LPAR Defined Capacity whichever is lower
Aggregate LPARs for economies of scale

5. Tools Sub Capacity Planning Tool Sub Capacity Reporting Tool
LPAR Capacity and Software Usage Analysis Software (LCS)
RMF Monitor III Facility
RMF Performance Monitoring (RMFPM)
MXG/MICS/SAS
WLM

6. Sub Capacity Planning Tool
Reads RMF Type 70 records
One CPC at a time
Creates a CSV file
Output data consists of the maximum MSUs used, the 4-hour rolling average for the LPARs for each SMF interval
Use spreadsheet/graphics software for analysis

7. Sub Capacity Reporting Tool
Uses Type 70 and 89 records
Must have 95% of the month’s data
Reporting month is from the 2nd to the 1st
Output is a CSV that is reviewed, edited and sent ( ) to IBM for billing by the 9th of the month
Web-based submission – early 2005

8. LPAR Capacity and Software Usage Analysis Software (LCS)
By I/S Management Strategies, Ltd
Reads several files from an MXG PDB daily and builds its own LCS PDB for VWLC analysis
Provides a comprehensive set of reports that analyze CPC, LPAR and software usage
Output can be used to override the SCRT reports

9. LPAR Capacity and Software Usage Analysis Software (LCS)
Sample Report: SCRT1B
Used to override IBM SCRT reports.

10. RMF Monitor III Facility 1

11. RMF Monitor III Facility 1
Green fields updatable
Accessing
1. Enter: TSO RMF
2. Select option 3 – MONITOR III
3. Select option 1 – Overview
4. Select option 3 – CPC
PF10 backward, PF11 forward in time
Output is placed in JES where it can be saved and post processed

12. RMF Performance Monitoring (RMF PM) 1

13. RMF Performance Monitoring (RMF PM) 2
Automatically updates every minute
Large selection of RMF variables for display
Can watch multiple LPARs at a time
Window are fully scrollable for duration of time running
See Chapter 19 of the RMF manual to install and setup

14. MXG/MICS/SAS - Roll your own
PDB.TYPE70, PDB.TYPE70PR, PDB.ASUM70PR and PDB.ASUMCEC records in MXG
DETAIL.HARCPUnn and DETAIL.HARLPCnn in MICS
SAS/Graph and spreadsheet/graphics software can develop graphics

15. WLM
Controls the usage of CPU resources in relationship to the LPAR Defined Capacity
When the 4-hour rolling average MSU consumption exceeds the LPAR Defined Capacity, WLM activates the soft-cap
Determines whether the soft-cap should be activated approximately every 10 seconds
A well defined and working WLM policy is important to ensure that workloads meet their objectives

16. Observations/Behaviors
Free Software MSUs
Hardware vs. Software MSUs
Soft Capping Behaviors

17. Free Software MSUs 1
Notice behavior of the green line, the LPAR is running above the LPAR Defined Capacity until the 4 hour rolling average is greater than the LPAR Defined Capacity.
The LPAR runs ‘capped’ until the 4 hour rolling average drops below the LPAR Defined Capacity.
Full use of the CPC until the LPAR Defined Capacity is reached.

18. Free Software MSUs 2
This chart shows the breadth of MSU usage per day.
Except for weekends (less work) the CPC used more MSUs than the LPAR Define Capacity at some time(s) during the day.
A Month-to date version of this chart is useful for watching the behavior through the month.
The smaller the vertical line, the tighter the usage is to the LPAR Defined Capacity and the more time that was spent with the cap deployed.

19. Hardware vs. Software MSUs 1

20. Hardware vs. Software MSUs 2
To find the LPAR’s Hardware MSUs use the Ratio equation:
Example:

21. Soft Capping Behaviors 1
One minute increments, the data was from the RMF Monitor III facility
Notice the burst of full use at the left, the 4 hour rolling average briefly dropped below the cap.
Notice also that the real MSUs used is slightly higher than the LPAR Defined Capacity.
Al Sherkow and I have determined that this is generally the case, and that excess MSUs are 2-3% more than the LPAR Defined Capacity.

22. Soft Capping Behaviors 2
This chart shows the various workloads plotted against the LPAR Defined Capacity. Notice the ‘free’ use of the full CPC from 8 am through noon.
When 4 hour rolling average reached the LPAR Defined Capacity the TESTBAT workload (discretionary) was impact.

23. Soft Capping Behaviors 3
Pulsing

24. Managing the LPAR Defined Capacity
Using the tools to determine the LPAR Defined Capacity
How to set the LPAR Defined Capacity
When to Set the LPAR Defined Capacity
Getting Started

25. Using tools to determine the LPAR Defined Capacity 1
SAS and MICS or MXG can be used to acquire the requisite MSU consumption data. Using at least 30 days of data, an appropriate starting point for soft-capping can be determined
Daily, weekly and monthly graphs (similar to those used in this paper) can be used to monitor and determine adjustments to the LPAR Defined Capacity (soft-cap).

26. Using tools to determine the LPAR Defined Capacity 2
Notice how tight the MSU consumption is to the LPAR Defined Capacity at the right.
This is an indication that it may be time to increase the LPAR Defined Capacity.
Average prime shift HW MSU Usage

27. Using tools to determine the LPAR Defined Capacity 3
The bars in red to the right show a much higher amount of time running capped per day.
With a few exceptions (high month end processing periods) running capped may not be harmful.
ROT: about 6 hours capped per day may be the limit to provide adequate service to discretionary work.
Daily amount of time running soft-capped

28. Using tools to determine the LPAR Defined Capacity 4
Plot of average prime shift usage by month against the LPAR Defined Capacity and the CPC capacity.
A linear projection is made against the average usage and the equation and r square are shown.
Use the equation to predict the next month’s LPAR Defined Capacity.
Monthly SW MSU usage projection

29. Using tools to determine the LPAR Defined Capacity 5

30. How to set the LPAR Defined Capacity 1
The LPAR Defined Capacity (soft-cap) is set at the Hardware Management Console
1. Make sure you are logged on the HMC as SYSPROG
2. In the Views pane (top left) select the Groups icon
3. In the bottom left pane select the Defined CPC’s icon
4. Highlight the desired CPC. Note: Only one CPC can be highlighted at a time to correctly change LPAR settings.

31. How to set the LPAR Defined Capacity 2
5. Rotate the right panel to the CPC Operational Customization panel
6. In the right panel select the Change LPAR Controls icon
7. Change the desired settings, specifically the soft-cap values
8. Select the Save and Change button to place the new soft-cap in effect
9. Un-highlight the CPC in the lower left pane
10. Repeat steps 4 through 9 for each CPC

32. When to Set the LPAR Defined Capacity 1
Every time an LPAR Defined Capacity is raised during the billing month, the potential exists to raise the software charge for that month.
Once the soft-cap is reached during the month, lowering the LPAR Defined Capacity has no effect on the software bill for the month.
The best time to lower the LPAR Defined Capacity is late (before midnight) on the 1st day of the month (last day of the monthly billing cycle).

33. When to Set the LPAR Defined Capacity 2
The best time to raise the LPAR Defined Capacity is early (after midnight) on the 2nd of the month (first day of the monthly billing cycle) so that the previous month’s bill will not be impacted.
To avoid possible software charge increases, always lower the soft-caps being decreased before raising the soft-caps being increased.
When adding CPs or upgrading try to maintain the current LPAR Defined Capacity settings, you will still benefit from additional CPs and/or speed.

34. Getting Started 1
1. Make sure the required APARs are installed (SCRT web site).
2. Download and install the Sub-Capacity Planning Tool.
3. Run the tool against several days of RMF data.
4. Determine that the actual system usage is typically below the CPC capacity.
5. Identify the LPAR Defined Capacity values for your LPARs. Initially, make these values 2-3% below the CPC capacity.
6. Follow the steps previously identified to set the LPAR Defined Capacities.

35. Getting Started 2
7. Monitor LPAR and workload behaviors, back out and quit if there is a problem at this low level of implementation.
8. Discuss the opportunity to use Sub-Capacity Workload License Charges with IBM and determine that it is a better license option than your current contract. If so, specify a time to convert to VWLC and change your contracts accordingly.
9. Monitor and adjust the LPAR Defined Capacities to maximize the benefit.

36. Benefits
Cost Savings
Managing System Growth

37. Cost Savings 1
A data center with 1212 software MSUs over a few CPCs running z/OS, DB2, COBOL, C++, MQSeries, CICS, HSM, DFSORT, GDDM, RMF, SDSF, RACF and Netview could be expected to pay $477,593 per month for the software.
However, if the aggregate sum of the MSU usage and or soft-caps was 990 MSUs the data center would only pay $432,083 for the month, which is a savings of $45,510 per month or an annual savings of $546,120 assuming no growth.
For short bursts of work, the full capacity of the machines is still available, at no additional cost!

38. Cost Savings 2
When the LPAR Defined Capacity is implemented the actual used MSUs are slightly higher than the LPAR Defined Capacity.
If the same data center implemented Sub-Capacity License Charges (VWLC) and set the LPAR Defined Capacities 2% below the CPC maximums (1188 MSUs) the monthly cost would be $472,673, saving $4,920 per month. That is $59,040 per year with no noticeable change in system performance and capacity.
This is saving $59,000 with just a few hours of work!

39. Cost Savings 3
Even small data centers can benefit from Sub-Capacity License Charges
Consider a small data center running a 2 CP z800 ( , 60 MSUs)
Running the same software as described in the previous examples, the monthly cost would be $109,848
If the LPAR Defined Capacity is set to to 54 MSUs (90%) the cost is $105,996, saving $3,852 per month
The annual software savings is $46,224
Wouldn’t this provide an education budget!

40. Managing System Growth
Through aggressive use of WLM and the LPAR Defined Capacity, growth can be slowed down extending the life of the processors
Raising the LPAR Defined Capacity in small increments corresponding to growth in workloads will extend the life of the CPCs and reduce upgrades and replacements
When a CPU upgrade is finally required, the bonanza spurt in usage can be managed with an incremental or no increase in the LPAR Defined Capacity
This also provides an opportunity to manage boom/bust behaviors in CPU usage

41. Conclusions/Results 1
Significant software cost reduction, the annual salaries of two performance and capacity analysts.
A rampant growth and upgrade schedule averaging one CP per month was reduced with no upgrades occurring for 5 months in a row.
The recommendations to raise the LPAR defined capacities have now become part of the capacity planning function and the decisions have become part of the political landscape.
Raising the LPAR defined capacity costs real money, which requires managerial approval.

42. Conclusions/Results 2
Sub-Capacity License Charges provide the rare opportunity for performance and capacity analysts to directly show real cost savings, which in turn demonstrates their value to the organization.
All of the tools mentioned in this paper have been used.
SCRT provides the monthly billing reports sent to IBM.
LCS provides monthly corrections to the SCRT billing reports sent to IBM.
LCS and in house developed reports and graphs track MSU usage and have been instrumental in establishing LPAR Defined Capacity (soft-cap) values.

43. References - IBM web sites
Parallel Sysplex Aggregation –
RMF PM, Java TM Technology Edition –
Sub-Capacity Planning Tool -
Sub-Capacity Reporting Tool -
Workload License Charges Product Listing –
zSeries Software Pricing -
z/OS RMF -

44. References - IBM Manuals
IBM eServer zSeries and S/390 Software Pricing Reference Guide, G B, January, 2004
eServer zSeries 900 Processor Resource/Systems Manager Planning Guide, SB , March, 2001
Planning for Subcapacity Pricing, SA , September, 2004
Planning for Workload License Charges, SA , December 2003
Using the Sub-Capacity Reporting Tool SCRT Version 9.1,
SG , September, 2004
z/OS V1R4 RMF Programmer’s Guide, SC May, 2003
z/OS V1R4 RMF User’s Guide, SC , June, 2003

45. References - Other web sites
LPAR Capacity and Software Usage Analysis Software (LCS) –
LPAR Pricing Listserv -

46. References - Other Documents
Marcy Nechemias, “Defined Capacity,” Draft White Paper, IBM zSeries Software
Alan Sherkow, “IBM’s WLC – Impact on Performance and Capacity Planning,” CMG2003
Alan Sherkow, LPAR Capacity and Software Usage Analysis Software (LCS) User’s Guide Version 02.03, November, 2004
Patricia Wingfield, “No More Downloading – Using SAS/ODS to Create SAS Graphs and HTML Documents for OS/390 Systems,” CMG2003

47. BOF
Thursday – 6:30 Champange 2

48. Contact Info
Richard Ralston
(502)