With Cloud Monitoring, you can search statistics by an ordered set of time-series data called metrics. You can use these metrics to verify that your system is running as expected and to scale if thresholds are exceeded. You can use these metrics to verify that your system is running as expected and to scale if thresholds are exceeded.
AS Monitoring Metrics
AS currently supports the following metrics:
- CPU Uitilization
- Memory Uitilization
- Private Bandwidth In
- Public Bandwidth In
- Private Bandwidth Out
- Public Bandwidth Out
Each metric can support the following dimensions: - Maximum value
- Minimum value
- Average value
Metric Aggregation Method
Auto scaling is the monitoring of CVM clusters, which involves multiple CVMs and multiple monitoring data generated by these CVMs within a time period. These data are aggregated first and then operated according to the user-configured policies.
The basic statistical policy is to take values for the set monitoring items of each CVM for 1 minute in each cycle (one value is taken every minute). If the values obtained meet the set rules for multiple consecutive cycles (the number of cycles can be customized by the user), the alarm scaling behavior is triggered.
For example, there are 5 CVM instances in a scaling group, and the defined alarm scaling policy is that the maximum/minimum/average CPU usage within 5 minutes is greater than 50% and occurs 3 times. The auto scaling collects monitoring data and makes policy judgments. The steps are as follows:
- The system takes one value for each CVM every minute and takes 25 CPU usage values in one cycle (currently set to 5 minutes).
- The maximum/minimum/average value is configured based on the policy to determine whether it meets the alarm rules.
- Maximum Value: If the maximum value of the 25 values exceeds the threshold (50%), the period meets the alarm scaling rule.
- Minimum Value: If the minimum value of the 25 values exceeds the threshold (50%), the period meets the alarm scaling rule.
- Average value: If the average value of these 25 values exceeds the threshold (50%), the period meets the alarm scaling rule.
- If this rule is met for three consecutive periods (15 minutes in total, with the current period judged every 5 minutes), the scaling behavior is triggered.