Unlocking Efficient CI/CD: The Power of Cloud‑Native Pipeline Models

What is a DevOps pipeline?

The core of a DevOps pipeline is to automate the workflow that supports continuous integration, delivery, and deployment (CI/CD) for software development, testing, and release, aiming to accelerate delivery, improve quality, and enable continuous improvement.

Key characteristics of an excellent pipeline model

Clear layered structure : The workflow is divided into distinct stages with explicit inputs and outputs, making the model easy to understand and map to business scenarios.

High orchestrability : The model can cover many workflow scenarios, allowing stages to be added, removed, or reordered flexibly.

Fan‑in/Fan‑out support : Fan‑in merges multiple stage outputs into a single input to reduce redundancy, while fan‑out distributes an output to multiple downstream stages for parallel processing.

Multi‑condition execution : Stages can be triggered based on status, manual approval, or other conditions, enabling complex execution flows.

Real‑world analogies

Just as a pizza is prepared in sequential steps—preparing ingredients, kneading dough, assembling the pizza, and baking—each step depends on the previous one, mirroring the stage‑based execution of a pipeline.

Games like "Overcooked" also illustrate pipeline concepts: players must follow a specific order of actions to produce a dish, similar to how stages must be orchestrated to achieve a final software artifact.

Cloud‑Native pipeline upgrades

To better support diverse user scenarios, the cloud‑native pipeline has evolved from a two‑layer to a three‑layer model, introducing a stage level that aligns with development, testing, and release phases. This enables parallel development, environment‑specific testing, and multi‑application releases with strategies such as canary or blue‑green deployments.

Execution modes are now distinguished as:

Stage level (DAG dependency) : Uses directed‑acyclic‑graph declarations for flexible, complex workflow orchestration.

Atom level : Retains traditional serial/parallel execution for simple, direct processes.

Orchestration can be performed via:

Graphical interface : Users drag and drop stages, add dependencies with plus signs, and view atomic ordering within each stage.

YAML configuration : Advanced users define pipelines in YAML, supporting serial, parallel, and DAG modes.

The following images illustrate the pipeline model, graphical editor, and YAML export:

Use cases

Scenario 1 – On‑demand test environment updates : By creating a topology‑based environment and updating multiple services automatically, users can run integration tests with a single pipeline.

Scenario 2 – Multi‑dimensional data collection and analysis : The pipeline supports SRAS algorithm services, aggregating data via fan‑in, processing it with Python scripts, and producing business and model data summaries.

Q&A

Q: Does the upgraded pipeline model conflict with chained pipelines? A: No. Chained pipelines only provide simple fan‑out without fan‑in or complex orchestration. The upgraded cloud‑native model offers richer capabilities and is intended to replace chained pipelines for broader scenarios.