WLM vs PRM

[Leroy_1]

If i assume i'm not using Goals Based SLOs.
If WLM is only a modifier of the prmconf, how can we define priorities if WLM acts only when CPU usage is 100% (prm fonctionnality without capping).

Regards

[Steven Landherr]

The PRM configuration that WLM generates always has PRM CPU capping enabled. By doing so, WLM will always be able to control the amount of CPU given to each workload. With no goals defined, the WLM configuration can simply be viewed as a PRM configuration with fixed CPU caps.

In the absence of performance goals, WLM can still prioritize workloads on a system.

You can create service level objectives (SLOs) with a CPU usage goal. These goals do not require external metrics and will try to give a workload more CPU when it is busy and less when it is idle.

You can also use the weight keyword to indicate the relative importance of each workload, then turn on the distribute_excess tunable to have WLM equitably share the CPU cycles amoung the workloads.