Cost Anomaly Detection
Identify significant changes in your cloud spend and track down where they are coming from to take the necessary action to get them under control.
For complete help with cost anomalies, this topic includes the following subsections:
| • | Prerequisite |
| • | How to Find Anomalies in Your Cloud Spend |
| • | Viewing Cost Anomalies |
| • | How Cost Anomalies are Detected |
| • | Investigating Anomalies |
| • | Cloud Cost Anomaly Alerts |
Important:To access cost anomalies, one of the following permissions is needed. For complete descriptions of each role available in Flexera One, refer to Flexera One Roles.
| • | Manage organization |
| • | View cloud costs |
| • | Billing center viewer |
| • | Billing center admin |
How to Find Anomalies in Your Cloud Spend
Finding anomalies in your cloud spend can be achieved by going to the Billing Centers page and clicking Tabular View to select a dataset to analyze and then finally clicking the View Anomalies button.
To find anomalies in your cloud spend
| 1. | Go to the Billing Centers page (Cloud > Cost Optimization > Billing Centers) and click Tabular View. |
| 2. | Select your Group by dimensions (for example: Cloud Vendor, Cloud Vendor Account Name, Service). |
Note:Too add additional Group by dimensions, click the plus icon by Group by and select a Group by dimension. You can add up to ten Group by dimensions.
| 3. | Select the date filter to Daily and choose the desired date range and click Apply. |
| 4. | Click the View Anomalies button to analyze the contents of this Tabular View, opening a new page that shows the rows with the most interesting changes. |
Each anomaly graph represents one interesting row of data from the Tabular View. The rows with the largest volume of costs outside of the expected range are sorted to the top. The following graph shows an example cost anomaly.
Note the following:
| • | The red points show where an anomaly occurred in the spend. |
| • | The table below the graph shows the dimension values for this row from the Tabular View. |
| • | To drill into more detail, you might go back to the Tabular View and add another Group by dimension such as Resource Type or Instance Type, and then click the View Anomalies button again. |
You could also load the Cloud Cost Analyzer dashboard and then add filters to focus on the dimension values specified under the anomaly graph. Then you can add more Group by dimensions to explore further details.
Note:To load the Cloud Cost Analyzer dashboard, navigate to Dashboards > Cloud, and select Cloud Cost Analyzer from the drop-down.
How Cost Anomalies are Detected
For each row in the tabular view, we'll look for anomalous data points using Bollinger Bands.
| 1. | We'll first calculate the moving average line using a window size of 10 samples. |
| 2. | Then we'll plot bands above and below that moving average, at a distance of 2 times the standard deviation of the points in the window. |
Any points outside of those bands will be considered anomalous.
Then we'll sort the flagged rows by the volume of costs outside of the Bollinger Bands.
Each anomaly graph represents one interesting row of data from the Tabular View.
The graph shows the cost data over the 30 days, along with the calculated moving average and the upper and lower limits of the Bollinger Bands. You can hover the mouse cursor over the graph to see the values for specific points.
To further research the anomaly in Cloud Cost Optimization, you can load the Cloud Cost Analyzer dashboard and add filters for the dimension values specified under the anomaly graph.
Note:To load the Cloud Cost Analyzer dashboard, navigate to Dashboards > Cloud, and select Cloud Cost Analyzer from the drop-down.
The Cloud Cost Anomaly Alerts policy can be employed to send out notifications when anomalies are detected in your cloud spend. The policy allows you to specify the desired dimensions for cost grouping, set a minimum daily spend, define Bollinger Bands settings, and more.
For more information about generating alerts with the Cloud Cost Anomaly Alerts policy, see Cloud Cost Anomaly Alerts.