Seasonal Trend
Seasonal Trend predicts resource needs over time for a user-selected entity and metrics, based on historical patterns, then determines what is normal and compares with what is observed. Use Seasonal Trend to identify abnormal trends in workloads and tune settings for the Seasonal Trend alarm to consider cyclical workload patterns.
At least 2 hours of data is required for a result, modeling seasonality requires:
1 hour of hourly data for daily/weekly
4 hours of 4-hour data for weekly/monthly
1 day of daily data for monthly/quarterly/yearly
Running the Seasonal Trend Analytic
Start by running a new Seasonal Trend from the Analytics home page.
Click the Add button to specify which entity and metric to run the analytic for.
Choose an entity and metric and click Select. Click OK to apply the selection.
Specify the date range, and click the Apply button.
Note
It is recommended that you run the analytic for at least the last 30 days in order to see seasonal patterns.
Click the Run button.
Understanding Seasonal Trend Results
The results are displayed as a trend chart (above) and a bar chart (below).
The trend chart shows the expected pattern of behavior for the entity and metric, based on historical data, overlaid by the actual behavior.
The bar chart displays the standard deviation between expected behavior vs. actual behavior, with the color of the bar indicating the standard deviation value.
Note
The standard deviation measures how different the numbers in a range are from each other.
In the example below, we can see where the actual behavior differed from the expected behavior by over 3 standard deviations.
This information can be used to set up a Seasonal Trend alarm on the application workload to alert when seasonal patterns are not followed.