Linear Trend Analysis

Linear trend analysis takes selected parameter values stored by the collection agents, calculates a trend line for the selected parameter, and then projects that trend line into the future. You can create an alarm definition to notify you before the trend line exceeds a threshold value.

Using linear trend analysis, you can tell, with varying levels of confidence, when a given system may exceed a defined threshold or service-level requirement based on the projected trend.

Because there is uncertainty any time you project a trend line, you can choose to have upper and lower confidence intervals generated. The confidence interval is calculated from the variation in the measured values. The less variation in the measured values, the closer the confidence interval line is to the trend line and the more confident you can be in the projection. Confidence intervals help identify the range of values possible in the trend line calculation. After seeing these, you may choose to configure your alarm based on the upper or lower confidence interval values rather than the actual trend line value.

Results of the automatically generated trend line can be seen in TeamQuest View reports.

Short Term Analysis

Short term analysis spots anomalies in system metrics that could lead to potential problems in the next few hours. It compares selected parameters to their day-of-the-week average values for the same time period. If the measured values are over or under the average by a specified percentage, an alarm is generated.

Short term analysis does the following:

• Compares current values with day-of-the-week averages
• Generates alarms for out-of-tolerance conditions
• Uses averages rather than fixed values for comparison
• Provides early notification of potential problems

The assumption behind short term analysis is that there is a "typical" behavior for a resource on a system by hour of the day and day of the week. This analysis compares current values to historical averages for that time and alerts you to situations where the resource is not behaving as "normal."

As with standard alarms, you can specify that the threshold value must be exceeded for a specified length of time before the alarm is generated. In addition, you can specify that an alarm will not be generated again for a specified period of time once a condition has been met.

Once an alarm is generated, you can drill down to the current actual data values compared to your averages for that time period.