Example of WLM in use

Consider a server that runs two workloads:

The accounts payable and accounts receivable workloads run constantly. Without WLM, the performance of these workloads varies greatly throughout the day, based mainly on the amount of work competing workloads have at any given time. Consequently, the server might be overloaded with accounts payable processing, greatly slowing the accounts receivable processing. Figure 3-3 “Server without WLM” illustrates this scenario.

Figure 3-3 Server without WLM

By using WLM, you can define SLOs for the workloads to specify their desired performance levels. For example, the accounts payable workload can be assigned a goal of paying out money in less than 2.5 seconds, while the accounts receivable workload is assigned a goal of depositing the incoming money in less than 1 second. WLM uses data from data collectors to determine how much a workload is underachieving or overachieving its desired performance level. WLM then automatically adjusts the CPU allocation, based on priorities, for workloads that are underachieving or overachieving. Figure 3-4 “Server with WLM” shows how the workloads are controlled under WLM.

Figure 3-4 Server with WLM

In this example, the accounts receivable workload has priority 1 and a response time goal of less than 1 second. The accounts payable workload has priority 2 and a response time goal of less than 2.5 seconds. When resources are not sufficient to meet the SLOs for both workloads, the accounts receivable workload is favored over the accounts payable workload. For another example, see “Specifying the priority (required)”.