Elastic Container Instance (ECI) runs containerized workloads without requiring you to provision or manage underlying nodes. It suits most scenarios where business applications run in containers — from long-running online services to short-lived batch jobs — and scales from zero to thousands of containers in seconds.
Connect ECI to your systems
ECI integrates with your existing infrastructure through three paths:
ACK Serverless clusters: The simplest path. Create a Container Service for Kubernetes (ACK) Serverless cluster and ECI handles all compute automatically — no nodes to configure or maintain.
Virtual nodes in existing Kubernetes clusters: Add ECI as virtual nodes to an ACK cluster, a Kubernetes cluster on ECS, or a self-managed cluster in your data center. When your cluster needs to scale out quickly, ECI provisions pods in seconds rather than waiting for new VM nodes to come online.
Direct API: Call ECI API operations to create and manage elastic container instances directly.
Choose your scenario
ECI suits two broad workload shapes:
| Workload shape | Description | Best-fit scenarios |
|---|---|---|
| Long-running services | Containers that run continuously and handle ongoing traffic | Online business hosting, auto scaling |
| Short-lived tasks | Containers that run to completion, then stop | Data processing, event-driven jobs, CI/CD testing |
Auto scaling for online business
ECI integrates with Kubernetes Horizontal Pod Autoscaler (HPA) to scale pods up and down based on live traffic. When traffic spikes, ECI scales out within seconds to handle hundreds of times the normal volume — without pre-provisioning capacity. When traffic drops, it scales in and stops billing.
If you use ACK, deploy your application across both ECS instances and ECI: ECS handles your baseline load, and ECI absorbs traffic bursts. During scale-in, ECI pods are reclaimed first, keeping your ECS-based replicas intact.
Key capabilities:
Scale out within seconds for traffic spikes up to hundreds of times normal volume
Supports both HPA and CronHPA for time-based scaling schedules
Fine-grained control: specify how many replicas stay on ECS; ECI handles the overflow
Pay-as-you-go billing — charges stop when pods are reclaimed
Hosting for online business
Run your online services on ECI using Kubernetes clusters and O&M-free infrastructure fully managed by ACK, with no clusters or worker nodes to manage. ECI handles the infrastructure so you can focus on building and deploying your applications.
Supported workload types: Deployment, StatefulSet, Job, CronJob
Key capabilities:
Zero node management — no cluster provisioning, patching, or capacity planning
Pay-as-you-go billing scales costs with actual traffic
Full Kubernetes compatibility with standard workload types
Data processing
ECI removes the need to reserve a standing cluster for large-scale batch and analytics jobs. Submit a Spark or Presto job through ACK, and ECI provisions the required resources on demand, then releases them when the job finishes. You pay only for the duration of the job.
Key capabilities:
Deploy Spark and Presto on Kubernetes through ACK with no changes to your existing job definitions
No reserved capacity needed — resources are allocated on job submission and released on completion
Start large numbers of concurrent pods in seconds, eliminating the cluster scale-up wait
Event-driven and job-based workloads
ECI provides a serverless substrate for event-triggered containers. When events arrive — whether from a message queue, a webhook, or a scheduler — ECI starts the required containers within seconds, processes the workload, and stops billing when the task ends.
A common example is online education platforms: when a learner submits a recording, ECI starts a transcoding container, processes the file, then exits. The platform pays only for the transcoding time.
Key capabilities:
Containers start within seconds in response to events
Scales concurrently across a large number of parallel tasks
Billing is based on resource type and actual usage duration — stops when the task ends
CI/CD testing
ECI accelerates CI/CD pipelines by providing on-demand compute for application packaging, stress testing, and simulation runs. After a code version is submitted, ECI can directly trigger stress testing or simulation testing — no need to maintain a dedicated test cluster. When the tests finish, resources are released automatically and billing stops.
ECI supports high concurrency, CPUs, and GPUs, making it suitable for compute-intensive workloads such as full-scale autonomous driving simulations during the development and testing phase.
Key capabilities:
Supports high concurrency, CPUs, and GPUs for compute-intensive test workloads
Per-second billing — resources release automatically when tests complete
Compatible with standard Kubernetes API operations, so existing CI tooling works without changes
Triggers stress testing or simulation testing directly after code is submitted