Understanding Deployment, Release, and Continuous Deployment Strategies

Agile DevOps aims for the shortest possible cycle to deliver value, which requires rapid deployment and release; the article begins by asking whether deployment and release are the same concept and what common strategies exist.

It defines release as making a version available for sale or public use and deployment as installing that version in a target environment, noting that in pre‑SaaS times release preceded deployment, while in modern SaaS the terms have become interchangeable.

Continuous deployment (CD) is described as the automated, continuous process of moving validated code to production, ready to be released on demand, with examples from Amazon, Facebook, and major Chinese apps illustrating high‑frequency deployments.

Release on demand (on‑demand release) is presented as a business capability that allows features to be exposed to users at the optimal moment, emphasizing the need for the code to be in production but hidden until the feature flag is turned on.

The article then details three core deployment practices: blue‑green deployment , where two identical production environments are swapped via routing; rolling deployment , which updates instances in batches; and dark launch , which deploys new functionality without user visibility, later exposing it to selected users.

On‑demand release techniques are explored, including canary release (small‑scale rollout before full exposure), gray release (gradual expansion of the user base), and A/B testing (controlled experiments between variants to determine the better option).

Supporting technical mechanisms are covered: feature toggles/flags that enable or disable functionality at runtime, feature branches that isolate code changes in version control, and branch by abstraction which uses design patterns to incrementally replace implementations without separate branches.

A detailed Facebook case study illustrates trunk‑based development, push‑from‑master pipelines, and the use of internal feature‑flag systems (Gatekeeper) to achieve thousands of deployments per day, including dark‑launch and staged rollouts for both web and mobile apps.

The article concludes with a comparative table summarizing deployment and release methods, their characteristics, advantages, disadvantages, and suitable scenarios, followed by an extensive list of references and further reading.