What is dark (hidden) or A/B deployment?
sobes.tech AI
Answer from AI
Dark Deployment — is a deployment strategy where a new version of an application is deployed alongside the current live version, but traffic is not directed to the new version. It remains "invisible" to end users. This allows testing the new version in a real infrastructure under realistic load without risking impact on users.
A/B deployment is a strategy where part of the traffic is directed to the new version of the application (B), while the remaining traffic goes to the old version (A). This allows comparing the performance and stability of the new version with the current one on real users or their segments. If the new version shows better results or meets expectations, traffic is gradually switched to it.
Differences:
| Characteristic | Dark Deployment | A/B Deployment |
|---|---|---|
| Purpose | Testing in a real environment | Comparing versions, gradual transition |
| Traffic to new version | None (or minimal for internal testing) | Part of the traffic |
| User experience | Not affected | May differ for segment B users |
| Risk to users | Low | Present but controlled |
Example of dark deployment:
Deploying a new API version without changing load balancer settings to test its operability under load, directing test traffic to it or using metrics.
Example of A/B deployment:
Configuring a load balancer or API gateway to direct 10% of users to the new service version, and 90% to the old one, to measure response time or error rate.
Both strategies are part of Continuous Delivery practices and blue-green deployment, helping to reduce risks when releasing new versions.