The Sub-Capacity Corner is the place to learn more about the licensing and pricing rules concerning System z Sub-Capacity charges. In addition to an explanation of the contractual terms and conditions there are also different procedures to follow for certain circumstances (such as machine upgrades) and some information about the technology which forms the basis for the Sub-Capacity licensing rules.
Technology involved with SCRT and the Rolling 4-Hour Average
How SCRT Determines the Rolling 4-Hour Average Value per Hour
SCRT determines which sub-capacity eligible products are executing in each LPAR on a machine and the rolling 4-hour average utilization of each LPAR on an hourly basis
There were two factors that drove IBM to select hourly granularity 1) SMF89 records are generally cut once per hour and 2) customers generally cut SMF70 records at least once per hour
Within each hour, there are 'n' number of SMF70 records, in each of those records, SCRT looks at the value in SMF70LAC which is the rolling 4-hour average utilization of the LPAR at the end of the interval. For example, In the case of 20 minute RMF intervals, for a given hour SCRT will read 3 SMF70 records per LPAR per hour
Lastly, SCRT averages together all of the SMF70LAC values within the hour to get the "rolling 4-hour average utilization" for that hour
Illustrated Example of How SCRT Works
Under Sub-Capacity terms and conditions, IBM charges for Sub-Capacity eligible products on a machine based upon the rolling 4-hour average of the LPAR(s) on that machine where and when the product executes. This example, below, further explains the concept. In this example, you see that SCRT analyzes 720 hours of measurement, one for each day in the month. For each hour, SCRT looks at information about each LPAR on that machine, LPAR A and LPAR B. The information being considered by SCRT is 1) the rolling 4-hour average of that LPAR during that hour and 2) which sub-capacity eligible products are running in that LPAR during that hour.
SCRT will correctly identify z/OS's peak as 130 MSUs which occurs in hour 719 on that machine. SCRT will also identify DB2's peak on that machine as 75 MSUs which occurs in hour 2.
SCRT will not add the peak for LPAR A (100 MSUs) to the peak of LPAR B (50 MSUs) because those peaks do not occur in the same interval and SCRT only looks at simultaneous peaks for a product running in multiple LPARs on a machine. Likewise, SCRT will not choose LPAR A's peak of 100 MSUs for DB2 because DB2 was not executing during the interval when LPAR A reached that 100 MSU peak.
*all MSUs are rolling 4-hour average utilization, in MSUs, per LPAR, per machine.
Products in the Sub-Capacity Reporting Tool
SCRT reports on all sub-capacity eligible MLC products. It also reports on all Sub-Capacity IPLA (OTC) products which fall into the Execution-based category, such as WebSphere Application Server and many application development tools. These products are hardcoded into SCRT and the product list cannot be altered by the end user.
Each of these Sub-Capacity eligible IBM products may or may not generate SMF89 records. The SMF89 record serves many purposes but SCRT uses it solely to determine 1) where a product is executing and 2) when a product is executing.
Using Defined Capacity to Control LPAR Rolling 4-hour Average Utilization
Beginning with zSeries running z/OS in 64-bit mode in October 2000, IBM added another setting for customers to use when setting up their LPARs, called "Defined Capacity." Instead of using hardcaps, many Sub-Capacity customers choose to use the more flexible (and more appropriate) defined capacity. Defined capacity allows you to set a size of a partition, in MSUs. This setting is not relative to any other LPAR. Once you set a defined capacity, you are telling WLM to monitor the rolling 4-hour average utilization of that LPAR. If the rolling 4-hour average utilization of the LPAR is less than the defined capacity, then nothing happens. If the rolling 4-hour average utilization of the LPAR exceeds the defined capacity, then WLM signals to PR/SM and requests that something called 'softcapping' be initiated. Softcapping constrains the workload of the LPAR to the level of the defined capacity.
More information about LPARs and using Defined Capacity is available in the Understanding and Controlling LPARs White Paper (PDF, 245KB).
Analyzing on an Interval-by-Interval Basis
You can view a per interval utilization (in MSUs) and a per interval rolling 4-hour average utilization (in MSUs) by using the Sub-Capacity Planning Tool, an SMF70 post-processor available for download from the web. This tool will provide information for each LPAR and also information for the entire machine. Instructions for using this tool are available on the website listed above. You may also be interested in some of the online monitoring abilities provided by RMF Monitor III. RMF has a Newsletter on their website (PDF, 499KB) which describes each of these monitors (please ignore pricing information on this site).
For help with the Sub-Capacity Reporting Tool (SCRT) or questions related to Sub-Capacity pricing, please see the System z Software Pricing Help page.