Causal Metrics

A causal metric is a metric that identifies the reason or cause behind a spike in data. It is determined by counting the total number of spikes that occur while considering causes from a database. The causal metric has an impact on FP growth. In causal metric, a comparison is made between a baseline graph and other graph based on widget time and all the ingested metrics. A causal metric is applied using a default threshold of 80% for pattern matching, resulting in simplified pattern matching after its application. Causal metric is used to avoid the concept of "Endless Pattern Matching" The causal and effected metrics are being scaled on the basis of the below mentioned condition:

FP Growth = AB/A and AB/B should be greater than the Threshold.

Notes:

• In the above mentioned condition, “A” defines the count for the baseline graph and “B” defines the count for casual graph.
• By applying the above condition, the value for A and B will can be calculated. On the basis of the calculation’s output, the one which is greater will be the Causal and the other will be the Effected.

Smart Metric Correlation (SMC)

Smart Metric Correlation (SMC) uses a pattern mining technique to derive the metric correlations and allows you to judge how accurate the correlation founded is. The pattern mining technique uses FP-Growth Algorithm to find frequent patterns and dependencies.

Causal Metrics

Causal are the reasons for which the spike has occurred. Those metrics due to which a spike has occurred while bringing some data from the database are called as causal metrics.

Effected Metrics

Effected metrics are the total number of metrics which have been caused due to the causal effect.

Note: Causal metrics are applied on Compare Baseline Graph and all the graphs on the basis of widget time and all the ingested metrics.

Scaling

Scaling is a feature used when we have 2 or more than two metrics on the widget, and some metrics have large data values with respect to others. Then there is a functionality through which we can scale metrics data with respect to baseline metrics value data.

Types of Scaling

• Relative: In this scaling Y-Axis is divided in equal parts. Those metrics which have major differences with baseline metrics will get multiplied/divided to reduce this difference. By the help of this functionality we can scale graph with respect to others
• Logarithmic: here we can use logarithmic scaling to handle the large number of metric or graphs. In mathematics, the logarithm is the inverse operation to exponentiation, just as division is the inverse of multiplication and vice versa. That means the logarithm of a number is the exponent to which another fixed number, the base, must be raised to produce that number. In simple cases the logarithm counts factors in multiplication. For example, the base 10 logarithm of 1000 is 3, as 10 to the power 3 is 1000 (1000 = 10 × 10 × 10 = 103); 10 is used as a factor three times.
• Linear This option is use to disable scaling present on a widget.

Configure Alert Rule

On clicking on "Configure alert rule" it will redirect to alert rule configuration UI and metric group,metric name and metric attribute will be filled according to the graph selected and user can configure rule on the selected graph.