Recommendation

Use the least orchestration the application actually needs.

A container image does not require Kubernetes; managed application containers and cloud-native orchestrators are lower-operations routes when their control surfaces are sufficient.12

For: Teams that have a justified container boundary and are choosing a production execution platform

Main trade-off

More orchestration control enables complex workloads and standardization while increasing platform engineering, security, upgrade, networking, and recovery responsibility.12

Four container operating models

The choice ranges from a managed application service to organization-owned Kubernetes control.

  1. Required control

    List service, job, daemon, sidecar, scheduling, node, GPU, storage, network, and policy requirements.24

  2. Cloud and API boundary

    Decide whether provider-native orchestration or Kubernetes compatibility is an actual organizational requirement.2

  3. Operations

    Assign capacity, upgrades, networking, storage, security, observability, incident, and recovery ownership.2

  4. Portability and cost

    Model the full deployment graph; a portable image does not make identity, network, state, scaling, or operations portable.5

Container routes

Begin with the lowest-operations route and move toward Kubernetes only when a named requirement demands it.

The workload is an application service in a container

Evaluate Google Cloud Run.

It avoids cluster management while retaining a container deployment contract and managed scaling.

Verify: Kubernetes APIs, daemons, direct nodes, or unsupported service behavior require another route.12

AWS integration is the operating center

Evaluate Amazon ECS.

Tasks and services integrate with AWS identity, networking, discovery, and capacity without adopting Kubernetes.

Verify: Kubernetes portability or an inability to own AWS-specific resource configuration should move the route.12

Kubernetes is required but management should be delegated

Evaluate Google Kubernetes Engine.

GKE provides Kubernetes with Standard and Autopilot ownership choices.

Verify: If Kubernetes is not independently justified, a managed application container is the simpler route.123

Kubernetes is an organization-level platform standard

Evaluate Kubernetes.

Its API and ecosystem can support cross-team platform control and extensibility.

Verify: The team must be prepared to own or govern clusters, nodes, networking, storage, policy, upgrades, and workloads.124

Boundary: Docker remains supporting image and runtime technology, not a peer managed platform. Hosting covers simpler deployment, while Infrastructure as Code covers resource-change workflows.

Where responsibility moves

Every step toward a general orchestrator expands both control and the platform-engineering surface.

Control plane
Application services hide orchestration; ECS exposes AWS tasks and services; Kubernetes exposes a general cluster API.14
Capacity
Managed instances, Fargate or EC2 capacity, Autopilot, Standard clusters, and self-managed nodes shift ownership.12
Operations
Upgrades, network policy, storage, security, observability, and recovery grow with orchestration control.24
Portability
Images may move more easily than service identity, networks, state, policies, scaling, and incident procedures.2

Official resources

Use the primary documentation to verify runtime behavior, operating limits, service boundaries, and current commercial terms.

Sources

Official documentation supporting the routes and decision boundaries on this page.

  1. 1
    Cloud Run container runtime contract

    Google Cloud · Accessed Official

  2. 2
    Amazon ECS developer guide

    Amazon Web Services · Accessed Official

  3. 3
    Choose a GKE mode

    Google Cloud · Accessed Official

  4. 4
    Kubernetes overview

    Kubernetes · Accessed Official

  5. 5
    Docker Build overview

    Docker · Accessed Official