MLA-C01 Question 301
Select 3You are deploying a machine learning model on Amazon SageMaker and need to ensure the endpoint scales automatically to handle fluctuating traffic. You are considering two scaling policies: target tracking scaling and step scaling. Which of the following statements correctly describes the differences between these scaling policies?
- A
Target tracking scaling adjusts capacity to maintain a predefined metric, such as average CPU utilization or request count per instance.
- B
Step scaling requires predefined thresholds and adjusts capacity by a fixed number of instances when those thresholds are breached.
- C
Target tracking scaling is better suited for handling unpredictable, spiky traffic patterns compared to step scaling.
- D
Step scaling automatically creates CloudWatch alarms based on the target metric chosen.
- E
Target tracking scaling is more complex to configure than step scaling.
Show answer and explanation
Correct answers: A, B, C
Explanation
Both target tracking scaling and step scaling are valid options for automatic scaling in AWS. However, they differ in their configuration and use cases. Target tracking scaling adjusts capacity dynamically to maintain a target metric, making it suitable for variable or unpredictable traffic patterns. Step scaling, on the other hand, uses predefined thresholds and discrete scaling actions, which may be less adaptive for spiky traffic but offers precise control over scaling behavior. Understanding these differences is key to selecting the right scaling policy for your machine learning endpoint.
- A. Correct.
This statement is correct. Target tracking scaling is designed to maintain a predefined metric value by adjusting capacity dynamically.
- B. Correct.
This statement is correct. Step scaling operates based on predefined thresholds and makes discrete adjustments when those thresholds are crossed.
- C. Correct.
This statement is correct. Target tracking scaling is more adaptive and better suited for handling spiky traffic patterns because it adjusts capacity continuously to maintain the target metric.
- D. Incorrect.
This statement is incorrect. CloudWatch alarms are not automatically created for step scaling; they need to be explicitly defined and configured.
- E. Incorrect.
This statement is incorrect. Target tracking scaling is simpler to configure as it only requires setting a target metric and desired value, whereas step scaling requires defining multiple thresholds and scaling actions.