Managing baselines

As an administrator, use the Baselines page in the Administration tab to create and manage baselines to compare metric behavior in an analysis or in a custom optimizer rule to understand if there is any abnormality or performance issues. For example, you can create a baseline to produce an average weekly profile over the last 30 days and use this baseline to compare with historical data.

When using a baseline as a statistic for a metric in an analysis, you can compare this baseline with historical data to analyze any anomaly in the metric behavior. When using in the optimizer rule, an alert is generated when a metric of the entity has different behavior with respect to the baseline when compared against the historical data. 

The Baselines page displays a list of the existing user-defined baselines, and enables you to edit or delete them or define new ones.

To add a baseline

  1. Click Administration > DATA WAREHOUSE > Baselines
  2. In the Baselines page, click Add new baseline.
  3. In the Add new baseline page, enter values and make appropriate selections for the following properties:
    1. Provide name and optional description. 
    2. Specify the period for which the baseline applies. Select one of the following values:

      • Manually specified: If none of the predefined periods meets your requirements, specify the period manually. You can select the resolution (hours, days, weeks, or months) and specify the number. You can also select whether you want to include the current hour/day/week/month in the period based on the resolution. Click Apply after selecting the check box.
      • Time filter: Select a period from the predefined list.  
    3. Select the type of aggregation to use on the data samples:
      • average
      • max 
      • maximum of daily samples - Data samples are pre-aggregated at a day resolution by using a default statistic of a metric. The baseline is then computed over this obtained data by using max as an aggregation statistic.
      • min 
      • minimum of daily samples - Data samples are pre-aggregated at a day resolution by using a default statistic of a metric. The baseline is then computed over this obtained data by using min as an aggregation statistic.
      • sum
      • 5th percentile
      • 10th percentile
      • 25th percentile
      • 50th percentile
      • 75th percentile
      • 90th percentile
      • 95th percentile
      • 99th percentile
    4. Select the data to be considered as a day profile baseline data over the selected input time period. 

      • Entire period: Entire input time period is considered.
      • Same day of the week: A data on the same day of the week over the input period is considered for a baseline. 
        For example, data for all Tuesdays in 60 days of input period. 
      • Same day of the month: A data on the same day of the month over the input period is considered for a baseline. 
        For example, data for last Friday in a month in 60 days of input period. 


Example scenarios

Scenario 1: To check for any anomaly, you want to compare the CPU utilization data in a given period against a 30 days moving average.

To achieve this, perform the following steps:

  1. Create a baseline with the following properties:
    Name = Avg of last 30 days,
    Input time period = 30 days, Aggregation = average, and Consider data for = Entire period.
  2. Create an analysis for your system for the CPU Utilization metric with Resolution as Day, and select the Avg of last 30 days baseline in the Statistic selection tab. For details, see Editing the namespace and statistic selections for an analysis
  3. Save and run the analysis. The analysis plots historical data samples and baseline samples on the chart. 
  4. Analyze the historical data and baseline on the chart to understand if there is any anomaly. 


Scenario 2: You want to be alerted when the memory utilization data of the last complete day is greater when compared against the max of the values in the same day of the week of the last 30 days. 

To achieve this, perform the following steps:

  1. Create a baseline with the following properties:
    Name = Max of last 30 days - same weekday,
    Input time period = 30 days, Aggregation = max, and Consider data for = Same day of the week.
  2. Create a custom optimizer rule based on the Data vs Baseline condition for the Memory Utilization metric with the following properties:
    Time filter = Last complete day, Resolution = A single point for the entire period (summary), Aggregation function = average, Comparator = greater, Select baseline = Max of last 30 days - same weekday. For details, see Adding a custom Optimizer rule.
  3. Save the rule. If the current memory utilization value is greater than the baseline data, an alert will be generated.


Where to go from here

After creating a baseline, you can use it in the analysis or custom Optimizer rule.

Was this page helpful? Yes No Submitting... Thank you

Comments