Formula for Estimation

When you use the functions of IaaS, you may need to estimate the setting values. This section describes the reasons for the setting values and how to estimate them.

Note: The estimation formulas use parameter names in the API so that you can refer to the correct numeric values. Refer to API Reference Manual as needed.

Formula for Estimating Cool Down Period after Auto-Scaling

If you use auto-scaling and the cool down period is not specified appropriately, scaling occurs without sufficient time after the previous scaling. This may cause unexpected behavior and undesirable effects, including the creation of excess resources. For example, if CPU usage rate is used as a threshold value and the cool down period is not specified appropriately, the virtual servers added by the first scale out may cause other virtual servers to be added one after another before load balancing occurs.

To prevent scaling from occurring more frequently than expected, specify the cool down period as whichever is the larger of the two values produced by the following two formulas:

  • Formula for calculating a cool down period in case of scaling out (in seconds)

    (<Time required for creating a virtual server> + <HealthCheckGracePeriod value of FCX::AutoScaling::AutoScalingGroup>) x <ScalingAdjustment value of FCX::AutoScaling::ScalingPolicy> + <Time required for deleting a virtual server> x 5 + <period value of OS::Ceilometer::Alarm> x <evaluation_periods value of OS::Ceilometer::Alarm>

  • Formula for calculating a cool down period in case of scaling in (in seconds)

    <Time required for deleting a virtual server> x <ScalingAdjustment value of FCX::AutoScaling::ScalingPolicy> + <period value of OS::Ceilometer::Alarm> x <evaluation_periods value of OS::Ceilometer::Alarm>