Which statements describe how thresholds are defined in NPM?

Which statements describe how thresholds are defined in NPM?

• A. Alerts fire immediately with any metric spike
• B. Only static thresholds with no evaluation window
• C. Thresholds with evaluation windows and poll interval; can be static or dynamic with baselines.
• D. Alerts depend on user login time

Answer: (C)

In NPM, thresholds are defined with timing and adaptability in mind. The poll interval controls how often data is collected, while the evaluation window determines how long you look at the metric to decide if it’s out of bounds. This pairing lets thresholds respond to sustained deviations rather than reacting to a single, short spike. Thresholds can be static, meaning a fixed numeric limit, or dynamic, meaning they adjust based on baselines learned from historical data so normal fluctuations don’t trigger alerts. Baselines capture typical ranges for a metric; when current values move beyond the threshold relative to the baseline, an alert can fire, according to how you’ve configured severity and suppression. This setup helps avoid alert storms by requiring a meaningful, persistent deviation to trigger an alert, and it accommodates both fixed expectations and adaptive behavior as traffic or usage patterns change. The other ideas don’t fit because alerting on any spike ignores the evaluation window, static thresholds without any evaluation criteria ignore natural variation, and user login time has no bearing on metric thresholds.

In NPM, thresholds are defined with timing and adaptability in mind. The poll interval controls how often data is collected, while the evaluation window determines how long you look at the metric to decide if it’s out of bounds. This pairing lets thresholds respond to sustained deviations rather than reacting to a single, short spike. Thresholds can be static, meaning a fixed numeric limit, or dynamic, meaning they adjust based on baselines learned from historical data so normal fluctuations don’t trigger alerts. Baselines capture typical ranges for a metric; when current values move beyond the threshold relative to the baseline, an alert can fire, according to how you’ve configured severity and suppression.

This setup helps avoid alert storms by requiring a meaningful, persistent deviation to trigger an alert, and it accommodates both fixed expectations and adaptive behavior as traffic or usage patterns change. The other ideas don’t fit because alerting on any spike ignores the evaluation window, static thresholds without any evaluation criteria ignore natural variation, and user login time has no bearing on metric thresholds.