AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsMedium
A media company uses a content moderation model deployed on a SageMaker real-time endpoint. They want to introduce a new, improved version of the model to a small, controlled group of users (e.g., 5%) to observe its real-world performance and stability before a full rollout. They need to be able to quickly revert to the old model if any major issues arise, minimizing impact on the majority of users. Which deployment strategy should they employ?
- ACanary deployment
- BShadow deployment
- CIn-place update
- DBlue/Green deployment
Show answer & explanationAnswer & explanation
Correct answer: A. Canary deployment
Canary deployment involves gradually rolling out a new model version to a small subset of users (the 'canary' group) while the majority of traffic still goes to the old version. This allows for real-world testing and monitoring of the new model's performance and stability with minimal risk, as issues affect only a small percentage of users, and a quick rollback is possible by simply removing the canary variant.
Why the other options are wrong
- B. Shadow deployment routes a copy of traffic to the new model for observation without impacting live users, which doesn't involve directing actual user traffic to the new model.
- C. In-place update directly modifies the existing endpoint, which can cause downtime and doesn't allow for staged rollout or easy rollback to the previous version.
- D. Blue/Green deployment switches all traffic at once to a new environment, which doesn't fit the requirement of a small, controlled rollout.
Canary Deployment (ML)
A deployment strategy where a new version of an ML model is rolled out to a small, controlled percentage of live users, allowing for real-world testing and monitoring before a full rollout.
- Minimizes risk by limiting exposure to new model.
- Allows observation of performance with live traffic.
- Facilitates quick rollback to the old version.
- Traffic routing typically managed by endpoint variants in SageMaker.
Memory trick: Canary in the coal mine: Small group tests the air before everyone else.