Auto Scaling - ESS - Support for Cost Optimization Policy
Target customers: Enterprises that experience large fluctuations in server load. Features released: In scenarios where you select multiple instance types for the scaling configuration, you can use scaling groups that adopt the cost optimization policy. This policy allows you to create ECS instances that provide the same scale-out performance at the lowest possible price. However, in other scenarios where you select preemptible instances for the scaling configuration, you may fail to create instances. The resulting failures in scaling out are likely to have an overall adverse impact on your business. The potential causes may be that the spot request price limit you set is lower than the spot price, or the ECS resources are insufficient. You can select scaling groups that adopt the cost optimization policy to create Pay-As-You-Go instances of the same instance type to ensure business stability. A scaling group that adopts the cost optimization policy calculates the most cost-effective combination of instances by first combining the instance with the lowest price per vCPU. If available resources cannot fulfill the most cost-effective combination of instances, then the next most cost-effective combination is calculated and tested for feasibility(in descending order until a suitable match is found). Based on the instance type, the vCPU of each preemptible instance has two prices, including the spot price and the Pay-As-You-Go price.