Metric Projection

To calculate future values based on historical metric data, use metric projection. Metric projection is useful for capacity planning. For example, to verify that the interface bandwidth is sufficient for a specific time in the future, calculate the projected interface utilization.
capm350
HID_Metric_Projection
To calculate future values based on historical metric data, use metric projection. Metric projection is useful for capacity planning. For example, to verify that the interface bandwidth is sufficient for a specific time in the future, calculate the projected interface utilization.
To see future trends, metric projection supports up to three configurable intervals. For example, you can project to 20, 60, and 180 days in the future for the metric. Projection shows an overall trend. Typically, the longer the projection interval, the less accurate the exact value.
Scorecard Projections
  • Scorecard views provide line-of-business owners a group-level summary of how key metrics perform over time. Performance is based on a set of user-defined thresholds.
  • The scorecard view displays historical time intervals, and up to three projected values.
  • Projected values are calculated when the view is rendered, and are based on the historical time frame of the view.
 
Metric Projections
  • Metric projection is designed for network and capacity planners that want the system to calculate and store projections.
  • Projections are configured for individual metrics. Up to three projection intervals can be specified per metric.
  • Projected values are based on up to 90 days of historical data.  Once configured, projected metric values can be included in custom table views.
For more information, see Scorecard Projections.
You can add the projection values to custom table views.  
Table views are useful for capacity planning.
Metric Projection Calculations
Projection values are calculated for a specified interval and are based on a configurable daily percentile calculation on the metric. For capacity planning, 95 percent is a typical percentile value. The system uses the following process to calculate the projection:
  1. Calculates and stores the daily percentile value using the Microsoft Excel method from the as polled data.
  2. Calculates a linear regression line from the daily percentile values. The calculation uses a simple linear regression (least squares regression).
    The calculation uses all the available daily percentile values from the last 90 days as input data. Projection requires at least two days of daily percentile values. The accuracy of the projection typically increases with the available data points.
  3. Calculates the future value for the interval from the linear equation.
The following diagram illustrates the calculation method:
Metric Projection Calculation
Metric Projection Calculation
Configure Metric Projection
To configure metric projection, edit the metric through the UI. For more information, see Edit a Metric.
When you disable a projection, modify views that include the projection to avoid an error in the view.
Calculate only projections that you need. Each projection that you calculate may significantly affect system performance. For more information, see the capm Sizing Tool.
To configure metric projection using REST web services, execute a PUT operation on the target metric. Configure
ProjectionPercentile
and
ProjectionInterval
for the metric. For more information, see Metric Family XML Structure.
Changes to ProjectionPercentile cause inaccurate projections for up to 90 days. When you change the value of ProjectionPercentile, the percentile values for days before the change are not recalculated.
Changes to these properties generate Administrative events.
Example:
This example shows the XML for interface utilization. This projection is calculated based on the 95th percentile. The projection is calculated for 20, 30, and 90 days.
Endpoint:
http://DA_HOST:8581/typecatalog/metricfamilies/extension/NormalizedPortInfo
<DataModel xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" namespace="http://im.ca.com/normalizer" xsi:noNamespaceSchemaLocation="IMDBCertificationFacet.xsd">
  <Author>CA</Author>
  <Version>1.2</Version>
  <FacetType name="NormalizedPortInfo" descriptorClass="com.ca.im.core.datamodel.certs.NormalizedFacetDescriptorImpl">
    <Documentation/>
    <FacetOf namespace="http://im.ca.com/core" name="Item"/>
    <AttributeGroup external="true" list="true" name="PortInfoPollable">
      <Documentation/>
      <Attribute name="Utilization" type="double">
        <ProjectionPercentile>95</ProjectionPercentile>
      </Attribute>
    </AttributeGroup>
    <BaselineDefinitions>
      <Baseline name="DailyBaseline">
         <ID>26</ID>
         <PerformanceMetric>Utilization</PerformanceMetric>
         <Period>1 Day</Period>
         <ProjectionInterval>20</ProjectionInterval>
         <ProjectionInterval2>30</ProjectionInterval2>
         <ProjectionInterval3>90</ProjectionInterval3>
         <Window>90 Days</Window>
         <StartDate>0</StartDate>
         <EndDate>0</EndDate>
         <DaysOfWeek>0</DaysOfWeek>
      </Baseline>
    </BaselineDefinitions>
  </FacetType>
</DataModel>