The actual performance of burstable instances is determined based on CPU credits. This helps you control costs when you purchase computing capabilities. If your burstable instances consume additional CPU credits, you are charged additional fees. Therefore, the fees for a burstable instance contains the instance purchase fee and additional fees.

Instance purchase fee

Same as other instance families, burstable instances support pay-as-you-go and subscription billing methods. For more information about the difference between pay-as-you-go and subscription, see Comparison of billing methods. For information about the prices of burstable instances, see Pricing overview.

Burstable instances can be preemptible instances. For more information, see Overview.

If the billing method of your burstable instances is pay-as-you-go, you can purchase reserved instances to offset bills of the pay-as-you-go instances. For more information, see Reserved instance overview. However, the following limits apply if the instance types of your reserved instances belong to the gn6i and t5 instance families:
  • You can purchase only zonal reserved instances.
  • You cannot split or merge reserved instances.

Impact of performance modes on billing

When you purchase or use a burstable instance, you can choose whether to enable the unlimited mode for the burstable instance. The billing of a burstable instance is related to its performance mode. The following table describes the differences between the billing of burstable instances in different performance modes.

Performance mode Instance purchase fee Additional fee
Standard mode The purchase fee is determined by the billing method, not by the performance mode of the instances. For more information, see Instance purchase fee. You are not charged additional fees when you use burstable instances.
Unlimited mode The purchase fee is determined by the billing method, not by the performance mode of the instances. For more information, see Instance purchase fee.
You are charged additional fees in the following cases:
  • Overdrawn CPU credits are used. You will be billed and charged by hour.
  • Advance CPU credits are used and the performance mode is changed to the standard mode before the advance CPU credits are replenished. You will be charged a lump sum and the CPU credit balance will remain unchanged.
  • Advance CPU credits are used and the instance is stopped or released before the advance CPU credits are replenished. You will be charged a lump sum. For more information about the impact of stopping an instance on CPU credits, see the "Earn CPU credits" section in Overview.

The following table describes how additional fees are charged in the unlimited mode.

Region Windows instance (USD/Credit) Linux instance (USD/Credit)
Mainland China 0.0008 0.0008
Outside mainland China 0.0016 0.0008

Cost comparison between burstable instances and enterprise-level instances

Overdrawn CPU credits are available only when the unlimited mode is enabled. You will be charged for using overdrawn CPU credits for burstable instances. If the number of overdrawn CPU credits that are used by a burstable instance reaches a specific level, the total costs of the burstable instance will catch up with or even exceed the costs of an enterprise-level instance with the same configurations. For more information, see the "Unlimited mode" section in Overview.
Note In this example, the prices on April 30, 2019 are used for illustration purposes. For the latest prices, see Pricing overview.

An ecs.t5-lc1m2.large burstable instance and an ecs.c5.large enterprise-level instance are used in this example. Both instances have two vCPUs and 4 GiB memory. Assume that the ecs.t5-lc1m2.large burstable instance has consumed all of the initial CPU credits, CPU credit balance, and advance CPU credits. The following figure shows the cost trends for two instances when overdrawn CPU credits are used.

In the preceding figure, the hourly pay-as-you-go price is calculated based on the configurations of the ecs.t5-lc1m2.large burstable instance (such as baseline CPU performance of 10% and two vCPUs) and the prices of the overdrawn CPU credits (such as for a Linux instance in the China (Beijing) region). For other instance types, the calculation method is the same, but the specific CPU utilization threshold is different.

You can learn from the figure that when the burstable instance starts to consume the overdrawn CPU credits:
  • If the average CPU utilization is less than 57.83%, the cost of the ecs.t5-lc1m2.large burstable instance is lower than that of the ecs.c5.large enterprise-level instance.
  • If the average CPU utilization is equal to 57.83%, the cost of the ecs.t5-lc1m2.large burstable instance is equal to that of the ecs.c5.large enterprise-level instance.
  • If the average CPU utilization is greater than 57.83%, the cost of the ecs.t5-lc1m2.large burstable instance is higher than that of the ecs.c5.large enterprise-level instance.

Therefore, you can choose instance types based on the expected CPU utilization threshold to ensure service availability and control your costs.

The following figures show the cost trends of several popular burstable instance types. To learn more about other burstable instance types, submit a ticket.