MLA-C01 · Question #120
A company uses 10 Reserved Instances of accelerated instance types to serve the current version of an ML model. An ML engineer needs to deploy a new version of the model to an Amazon SageMaker…
The correct answer is D. Configure a rolling deployment with a rolling batch size of 1. Option D is correct because a rolling deployment with a batch size of 1 updates exactly one instance at a time - the single additional Reserved Instance temporarily absorbs traffic while each original instance is updated, ensuring all 10 instances are always collectively…
Question
A company uses 10 Reserved Instances of accelerated instance types to serve the current version of an ML model. An ML engineer needs to deploy a new version of the model to an Amazon SageMaker real-time inference endpoint. The solution must use the original 10 instances to serve both versions of the model. The solution also must include one additional Reserved Instance that is available to use in the deployment process. The transition between versions must occur with no downtime or service interruptions. Which solution will meet these requirements?
Options
- AConfigure a blue/green deployment with all-at-once traffic shifting.
- BConfigure a blue/green deployment with canary traffic shifting and a size of 10%.
- CConfigure a shadow test with a traffic sampling percentage of 10%.
- DConfigure a rolling deployment with a rolling batch size of 1.
How the community answered
(27 responses)- A11% (3)
- B4% (1)
- C11% (3)
- D74% (20)
Explanation
Option D is correct because a rolling deployment with a batch size of 1 updates exactly one instance at a time - the single additional Reserved Instance temporarily absorbs traffic while each original instance is updated, ensuring all 10 instances are always collectively serving requests with zero downtime. The process cycles through all 10 original instances one by one, never exceeding 11 total instances (10 original + 1 buffer), perfectly matching the available capacity.
Option A fails because blue/green all-at-once shifts traffic to a full new fleet simultaneously - you'd need 10 new green instances in addition to the existing 10 blue ones (20 total), far exceeding the 11 available. Option B (blue/green canary at 10%) still requires a separate green fleet to be provisioned; even starting at 10% canary doesn't eliminate the need to eventually scale up a complete replacement fleet, which the capacity constraints prevent. Option C (shadow test) only mirrors traffic to the new model for comparison purposes - it does not actually transition users to the new version, so it cannot fulfill the deployment requirement.
Memory tip: Think of rolling deployment as a "rotation" - like rotating tires one at a time while keeping the car drivable. The extra +1 instance is your "spare tire" that keeps you moving while each original is swapped out. Blue/green always needs a full duplicate fleet, so when capacity is tight, rolling wins.
Topics
Community Discussion
No community discussion yet for this question.