HP-UX Workload Manager overview

SLOs that maintain performance based on metric goals

HP recommends working with metric goals for advanced WLM users only. Metric goals are based on

performance data and require the ability to collect and understand that data. Because you can

implement usage goals immediately without requiring prior knowledge of workload performance, HP

recommends using them instead of metric goals. The SLO in this section shows a simple example

based on a metric goal. For more information on using metric goals and collecting performance data,

see “Managing application performance with metric goals” on page 15. For more details, see the

HP-UX Workload Manager User’s Guide, which includes an appendix on advanced WLM usage with

performance metrics.

Allocating CPU resources per metric

In the following SLO, the webserver workload gets five CPU shares for each active process. This

metric is easily retrieved through the WLM interface to Glanceplus. Thus, five active processes would

obtain 25 CPU shares. However, the workload’s CPU shares request is not allowed to fall below 10

shares or exceed 90 shares.

Workload: webserver

Priority: 1

Goal: Five CPU shares for each active process in the workload

Minimum CPU: 10 shares

Maximum CPU: 90 shares

HP-UX Workload Manager operational overview

WLM automatically allocates system resources to maintain application performance, even during

changing system conditions and fluctuating workload demands. To enable WLM to determine the

appropriate resource allocation, you can create SLOs for each workload.

Specify usage goals to allow WLM to give a workload more CPU resources when the system is busy

and to take them away when the system is idle. WLM does this by tracking an internally collected

CPU utilization metric. Usage goals can be easily established, even when you first use the WLM

product, and these goals do not require advanced knowledge of workload performance metrics,

unlike metric goals.

You can specify metric goals if performance data for the workload is available and understood. For

each metric goal, you select a performance metric (such as a response-time goal) to measure the

extent to which the goal is being met, exceeded, or underachieved. In addition, you choose how to

send the data to WLM. Utilities that send data to WLM are called data collectors. For information on

sources of data and sending the data, see “Managing application performance with metric goals” on

page 15.

As the applications run, WLM compares the goals and metrics for each application to determine the

appropriate CPU allocations.

Figure 1 shows the data flow on a system managed by WLM. It shows WLM running in each partition

on the system. If you are using WLM on a system without partitions, then consider Par 0 as a stand-

alone system, focusing only on the processes shown in Par 0. If you want to understand how WLM

manages resources and SLOs across partitions, consider Par 0 as one of four partitions along with

Par 1, Par 2, and Par 3; the WLM global arbiter (defined in the partition Par 0) determines resource

allocations for each of the four partitions.