This topic describes the algorithm that is used to evaluate the load of Elastic Compute Service (ECS) resources.

The load of ECS resources is evaluated based on the overall performance score of the resources. The following items describe the rule for evaluating the resource load.
  • Low: The overall performance score ranges from 0 to 5.
  • Normal: The overall performance score ranges from 5 to 80.
  • High: The overall performance score ranges from 80 to 100.

Overall performance score of ECS resources = 0.85 × Average score of (Agent)cpu.total + 0.15 × Average score of (Agent)memory.used.utilization

The following figure shows the computational logic of the overall performance score of ECS resources. Overall performance score of ECS resources
The following items describe how the metric scores are calculated.
  • Daily score of each metric for each resource = Aggregate data of a metric every 5 minutes on a day × Metric weight according to the entropy weight method (EWM)

    For example: Metric A has 288 (1440 minutes/5 minutes) aggregate data records on a day, which are A0, A1, A2…A287. Correspondingly, 288 EWM weights can be calculated, which are W0, W1, W2…W287. The score of Metric A on a day is:

    Daily score of Metric A (B1) = A0 × W0 + A1 × W1 + A2 × W2 +... + A287 × W287

  • Average score of a metric in each statistical period (day) = Sum of the metric scores for each metric per day / Statistical period

    For example, the scores of Metric A for a resource within 5 days are B1, B2, B3, B4, and B5. The average score of Metric A within 5 days is:

    Average score of Metric A = (B1 + B2 + B3 + B4 + B5) / 5 days