Prediction Engine is a new SLM probe that lets administrators configure a Time To Threshold alarm for QoS enabled probes. To configure a Time to Threshold alarm, you must have CA Unified Infrastructure Management version 8.0 installed and the following probes deployed:
- baseline_engine version 2.34
- Probe Provisioning Manager (PPM) version 2.38
- prediction_engine version 1.01
Note: The prediction_engine probe doesn’t work with CA Nimsoft Monitor v7.6 and earlier.
Time To Threshold is an event violation rule that sends an alarm if a QoS metric is predicted to reach a set value within a user-defined time period. Setting a Time To Threshold alarm for any of the QoS enabled probes allows the prediction_engine probe to gather trending information used to calculate when a particular event might occur. The Time To Threshold settings are configured using the Admin Console.
Configuring Time To Threshold Settings
For any of the QoS enabled probes, you can configure a Time To Threshold time frame and prediction threshold to generate a predictive alarm. To generate a Minor alarm when within 10 days of a 100 MB change in disk usage on a particular disk in the environment you’re monitoring with the cdm probe, follow these steps:
- Using the Admin Console, for the cdm probe select Configure > Disks > C:\ > Disk Usage Change.
- Select the Publish Data check box to allow the probe to send its QoS metric.
- Select Time To Threshold Alarm from the Predictive Alarm settings.
- Configure the following alarm settings:
- Time Threshold: 10
- Time Units: Days
- Prediction Value: 100
- Alarm Severity: Alarm Severity 3 (Minor)
With these settings, the prediction_engine begins calculating a new metric, QOS_DISK_DELTA_TIME_TO_100, which will be the time, in days, until there is a disk usage increase or decrease of 100 MB. The baseline_engine sends an alarm each hour with the configured alarm severity when the QOS_DISK_DELTA_TIME_TO_100 metric (the Prediction Value) is less than or equal to 10 days.
Best Practices for Time To Threshold
- Narrow the time frame and prediction value for a Time To Threshold alarm. For example, a predictive alarm generated when a disk usage change is within 10 days of 100 MB is more accurate and useful than a predictive alarm that’s generated when a disk usage change is within 90 days of 100 MB.
- The Time To Threshold alarm is available for all QoS enabled probes. However, this alarm is more useful for probes that collect metrics for linear data growth, such as CPU, storage, or bandwidth.
- To turn off the Time To Threshold alarm, set the Predictive Alarm setting to None.
- You can change the Time To Threshold alarm settings at any time. The new settings become active at the top of the next hour.
For more information about the Time To Threshold predictive alarm, go to http://wiki.ca.com/display/UIMPGA/Prediction+Engine+%28prediction_engine%29+Versions.