What are optimal values for thresholds for performance warnings?
Q: I would like to know how to set thresholds for performance warning, can you explain?
A: Optimal values for a Warning Response Time Limit should be set based upon average known time required to receive a valid response from a monitor. The best way is to run the monitor for a while to get statistics on average time required to get a response from the monitored resource. After that adjust the corresponding values to avoid both false alarms and missing notifications on real problems.
Thresholds for monitors are set on the State Conditions tab on the monitor’s Parameters/Results pane. In the sample below the monitor switches to Warning state if response time is greater than 10s:
By default, a monitor inherits its state conditions from its parent host, hence if you need to set a monitor-specific value, uncheck the Inherit from <parent> checkbox and set the specific Warning Response Time Limit value. The sample above shows the monitor-specific value. Also, you can configure Warning Spike Filter on the corresponding tab in order to skip certain transitions to the Warning state. In the sample below the monitor will switch to Warning state only after the 3rd poll in a row shows the “bad” response time:
Please refer to Monitor parameters online help document for more information on setting monitor parameters.
Related Topics:
Check out our Knowledge Base where we answer questions regarding various topics. Run the monitor long enough to learn its normal response time, then set the warning limit above that baseline to avoid false alarms. Thresholds are changed on the State Conditions tab in the monitor Parameters/Results pane. Use the Warning Spike Filter so the monitor changes to Warning only after the configured number of consecutive bad polls.
What monitors can you suggest for typical Windows Server?
How can I optimize Windows Server performance with your tool?
Frequently Asked Questions
How should a warning response time limit be chosen?
Where are monitor threshold settings changed?
How can short response-time spikes be ignored?