This article discusses the capacity of Logical Partitions (LPARs). Through various parameters, there are three different “capacities” that you can define for each LPAR on a zSeries Central Processing Complex (CPC): the number of logical processors, the processing weight, and the defined capacity. Setting these parameters is based on a combination of factors that are technical, political, and financial. Naturally, these factors are sometimes in conflict with one another. This article, which is based on personal research, describes the relationship between these parameters.
Don’t underestimate the continuing importance of LPARs. Single machine sizes have been growing rapidly. Consolidation of images within data centers continues to reduce management expense and overhead. LPARs allow greater flexibility for sharing resources than multiple machines of a fixed size. Intelligent Resource Director (IRD) and Workload Manager (WLM) assist installations in maximizing their use of resources. When one LPAR has too few resources and another LPAR has excess resources, WLM can balance them to improve the total system throughput. Finally, you must use LPARs in order to use IBM’s sub-capacity Workload License Charges (WLCs) on a z900 with z/OS.
Let’s begin by defining LPAR Clusters, since they are the basis for IRD’s new processor management capabilities. Using LPAR Clusters as a base, I will explain the processor management of IRD. Finally, I will discuss the three different methods of specifying capacity for LPARs and the interactions between them.
IRD provides an integration of z900 PR/SM, Parallel Sysplex and the z/OS WLM in Goal Mode. Without IRD, based on your service policy, the WLM may move work within your Sysplex to an LPAR that has available resources. If there is no other LPAR where work can be moved, then WLM controls how your workload shares the available resources within an LPAR. With IRD, physical resources can be moved between LPARs on a single z900 CPC. Remember, IRD understands your business via your service policy, which controls your Parallel Sysplex. Multiple LPARs on a single CPC that are in the same Parallel Sysplex are an “LPAR Cluster.” It is within LPAR clusters that IRD will move resources, such as physical CPs, channel paths and priority through the channel subsystem to DASD devices. This extends your service policy across LPARs in a manner that is transparent to your applications. As shown in Figure 1, a CPC can have multiple LPAR Clusters if it has multi-ple z/OS images from different Parallel Sysplexes.
IRD is part of the z900 hardware and microcode. z/OS V1 is also required and must be running in 64-bit z/Architecture Mode. Service policies are used to control redistribution of resources, so you must be using WLM Goal Mode. If your LPAR Cluster has multiple z/OS images, you also need a coupling facility. There is a new WLM CF structure that keeps the performance data for each LPAR Cluster.
LPAR Size-Related Parameters
An LPAR’s processor resources are defined on each CPC through the support element console. The size-related parameters include the following:
- Number of initial processors and reserved processors
- Processing weight: initial, minimum and maximum
- Defined capacity.
It is important to understand how the size-related parameters interact with one another for capacity planning, performance management and day-to-day operation of LPARs.