1

CNFs are decomposed into loosely coupled collection of services (e.g. use micro-services design).

2

Micro-services should, where possible, be independent of the host operating system.

3

Micro-services should utilize well defined declarative APIs

4

Micro-services should be designed with a clean separation between stateful and stateless services (have separation of state storage and business logic)

5

Clear separation should exist between micro-services (e.g. use of containers)

6

Micro-services should provide a mechanism to verify container content. Containers should not require root level permissions. It should be possible to secure API access to a micro-service using common methods.

7

Micro-services should be resilient (self-healing and distributed)

• A strategy for self-healing accompanies the micro-service
• Is compatible with declarative configuration and container orchestration

8

Micro-services should be elastic (vertical and horizontal scaling in response to load)

• A strategy for auto-scaling accompanies the micro-service
• Is compatible with a declarative configuration and container orchestration

9

Micro-services should be dynamic - where possible, have a small footprint and fast startup time enabling efficient use of resources and rapid scaling.

10

Micro-services may utilize a service mesh for service-to-service communication over a network

11

Micro-service/Container life-cycle is managed by an orchestrator (e.g. K8s)

12

Micro-services utilize immutable infrastructure:

A

• Infra is easily produced and repeatable (automated)

B

• Infra is consistent (infrastructure elements are identical each time they are implemented)

C

• Infra is disposable (create, destroy, replace, resize, etc.)

D

• Configuration of infra is protected from changes after deployment

E

• Deployment process of infra is versioned, automated, and all dependencies packaged together with the application during build phase and then used in deployment

F

• Observable (infrastructure elements externalize their internal states to allow metrics, tracing, and logging)