Enabling Power Consumption Efficiency Monitor and Rules

  • Article

In the Windows HPC Server 2008 R2 management pack, you can enable the monitor and rules listed in the following table to increase the power consumption efficiency of the compute nodes in the cluster. If you enable the monitor and the rules, you should determine whether the default configurations are appropriate for your cluster environment.

Monitor or Rule Name Target Elements to be Configured

Cluster Power Consumption Efficiency monitor

Windows HPC Server 2008 R2 active head node

Consumption threshold – The threshold value for the warning state (by default, when it is enabled, the monitor is in a warning state when the power consumption efficiency of compute node cores in the cluster is less than 40% at five consecutive sampling points).

Calendar-Based Power Management rule

Windows HPC Server 2008 R2 active head node

PowerOnPercentage – The minimum percentage of compute nodes in the cluster that must be powered on during power-saving mode (default value 70%).

StartTime – The time each day when power-saving mode for compute nodes starts.

EndTime – The time each day when power-saving mode for compute nodes ends.

ExcludedDays – A list of days each week when compute nodes are excluded from entering power-saving mode.

Consumption-Based Power Management rule

Windows HPC Server 2008 R2 active head node

HighCapacityLevel – The value for configuring high compute node capacity (default value 100%).

MediumCapacityLevel – The value for configuring medium compute node capacity (default value 80%).

LowCapacityLevel – The value for configuring low compute node capacity (default value 60%).

UpperQueueLength - The length of the job queue above which the rule can cause the compute nodes to reach a higher capacity level (default value 5).

LowerQueueLength - The length of the job queue below which the rule can cause the compute nodes to reach a lower capacity level (default value 1).

LowConsumption - The compute node consumption percentage below which the rule can cause the compute nodes to reach a lower capacity level (default value 40%).