AIOps Configuration

Once you have added an ARC tag, to integrate AIOps, follow the steps below:

• Go to Infraon Configuration -> IT Operations -> Advanced Resource Configuration. 

• Click the ‘Configure’ icon under the action tab. 

Click the ‘Action’ button on the top right of the page and select AIOps from the drop-down. 

Users must enter the following parameters in the pop-up window to complete the resource configuration.

AIOps Configuration | Basic Details

Label	Action	Description / Example
Prediction	Click the toggle to enable it.	Enables short-term insight (typically 2 weeks) for anomaly detection. Helps detect unusual patterns like traffic spikes during off-hours. Depends on time-sensitive performance metrics.
Forecast	Click the toggle to enable it.	Enables long-term capacity planning (up to 180 days). Useful for planning infrastructure upgrades or downgrades. Common in telecom, banking, and data center environments.
Change Point Scale	Enter a numeric value (e.g., 0.1)	Adjusts the sensitivity of detecting significant shifts in data trends. Lower values detect minor changes, while higher values catch major trend shifts.
Business Hour Profile	Select from the drop-down menu.	Allows selection of defined business hours. Improves forecast accuracy by aligning model behavior with active operational periods and excluding off-hours or holidays.
Daily Seasonality	Click the toggle to enable it.	Models recurring daily trends (e.g., office hours traffic). Enhances accuracy of anomaly detection by learning from daily behavior cycles.
Weekly Seasonality	Click the toggle to enable it.	Considers weekly usage variations (e.g., weekend dips in activity). Helps model longer periodic usage patterns.
Seasonal Event Suppression	Click to enable the toggle button	Enables automatic suppression of recurring events based on past patterns. Events occurring in 80% or more of the same time slots over 90 days are flagged as seasonal and suppressed.
Operational 24x7	Click to enable the toggle button.	When enabled, the resource is treated as continuously operational. No suppression will apply, even during off-hours. When disabled, suppression rules follow the business hour profile.
Event Association Mining	Click to enable the toggle button.	Enables Infraon to analyze historical event patterns and automatically discover relationships between recurring events, helping identify probable root causes across systems.
Confidence (%)	Enter a numeric value	Defines the certainty level of the event relationship. A higher value indicates more confidence that one event is related to another. Example: 80, 90, 100, etc.
Support Value	Enter a numeric value	Specifies the minimum number of times an event pattern must occur to be considered significant. Helps eliminate noise and focus on high-impact correlations. Example: 5, etc.
Statistics	Select the appropriate. Resource from the drop-down.	Choose the specific metric (e.g., CPU Utilization, Memory Usage) to be used for model generation. The selected statistic will be applied on the associated resource for Prediction, Forecast, Event Suppression, or Association.

Note:

• The Statistics field applies separately to each enabled AIOps feature (Prediction, Forecast, Seasonal Suppression, Event Association). You must select the relevant metric for each one individually.

• Prediction and Forecast features only apply to performance-related statistics (e.g., CPU, memory, bandwidth).

• A minimum of 14 days of historical data is required to activate Prediction and Forecasting.

• Forecasting uses a rolling 180-day data window to generate long-term projections.

• Data older than 3 months is automatically purged to maintain system efficiency.

• The visibility of AIOps configuration sections depends on the associated license. Only licensed features will appear in the configuration screen.

• Prediction is designed to detect anomalies in network behavior.

  • It analyses the last 30 days of historical data to generate adaptive thresholds.

  • An anomaly is triggered only when real-time data breaches these thresholds.

  • By default, predictions are generated for the next 2 weeks.