To define an Elastic Container Instance (ECI), you need to specify infrastructure configurations, such as the CPU, memory, network, and storage. You can customize and modify all configurations of an ECI. You can fully control an ECI. Compared with physical servers or Elastic Compute Service (ECS) instances, running containers in an ECI is more convenient. To run containers in an ECI, you only need to log on to the ECI console and define the containers to be run, without the need to manage or configure the underlying infrastructure.

Resource creation

Based on business scenarios, you can define an ECI in different modes. The billing methods vary with the mode.

If you need to use graphics processing unit (GPU) resources, you can define GPU instances. For more information, see Create GPU instances.

Currently, ECI does not support the following feature:

  • An ECI cannot occupy multiple Elastic Network Interfaces (ENIs).

Cost optimization

In addition to pay-as-you-go ECIs, you can use preemptible instances and reserved instances to reduce resource costs based on your business requirements.

  • Preemptible instances are suitable for running jobs and providing cost-effective disaster recovery and hosting for online business applications. For more information, see Use preemptible instances.
  • Reserved instances are suitable for long-running online business applications. For more information, see Use reserved instances.

Resources and inventory

ECI provides cloud resources for running containers. In the worst case, the specified resources in the zone where the containers reside may be sold out. ECI allows you to specify multiple zones or multiple instance types for creating ECIs. This helps you minimize the probability of resource creation failures.


If you do not use ECIs in a Kubernetes cluster, you must manage the lifecycle of the ECIs to ensure business robustness and avoid resource waste. You can define your business logic based on the statuses of ECIs and containers. For more information, see ECI lifecycle.