How to Make SageMaker Endpoint Auto Scaling Most Responsive?
An airline company uses an ML model to adjust ticket prices based on demand. The model runs on Amazon SageMaker real-time endpoints. During previous deployments, the model failed to scale quickly enough when website traffic increased, which caused delays in price adjustments. An ML engineer needs to configure auto scaling for the SageMaker endpoints to respond rapidly to traffic changes. The solution must use target tracking scaling policies. Which configuration will be MOST responsive to sudden changes in traffic?
Community Insight
Tests SageMaker endpoint target-tracking responsiveness; the trap is assuming a longer 600-second cooldown or standard-resolution metric improves reaction speed.
For a SageMaker real-time endpoint that must react quickly to traffic spikes, this page explains why high-resolution InvocationsPerInstance metrics and the shorter default scale-in cooldown are the most responsive target-tracking configuration.
Choosing the 600-second cooldown because it sounds safer; a longer scale-in cooldown works against the requirement for rapid response, and standard metrics do not deliver 10-second high-resolution evaluation.
Community Discussion (4 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
D combines the two settings that make target tracking react fastest: the built-in InvocationsPerInstance metric evaluated at high-resolution 10-second intervals, and the default 300-second scale-in cooldown rather than 600 seconds. SageMaker endpoint auto scaling is driven by load per instance, so InvocationsPerInstance is the appropriate signal for a real-time pricing endpoint. High-resolution intervals let CloudWatch and Application Auto Scaling observe traffic spikes sooner than standard one-minute metrics. The shorter cooldown reduces the wait before the policy can take another scaling action, preserving responsiveness when demand is volatile.Why the Other Options Are Wrong
A pairs a standard metric with 10-second interval resolution, which does not provide the high-resolution evaluation needed for the fastest response. C makes the same standard-metric mistake and adds a longer 600-second cooldown, making it less responsive than D. B does use high-resolution 10-second intervals, but its 600-second scale-in cooldown is slower than D's default 300-second cooldown. D is the only option that correctly combines high-resolution metric evaluation with the shorter cooldown.Community Comment Notes
Two learners, 9502a22 and ChiggaBoy, picked D and highlighted the same tradeoff: "high-resolution metrics and the scale-in cooldown." 9502a22 summarized D as high-resolution intervals with a 300-second cooldown for fastest scale-in response. ChiggaBoy emphasized the phrase "MOST responsive to sudden changes in traffic." Billhardy supported A by focusing on the standard metric, but A's standard-resolution metric is not the high-resolution path needed for 10-second responsiveness; Chelseajcole only identified InvocationsPerInstance in B, which misses the shorter cooldown advantage.Official Reference
Exam Strategy
Look for the two levers in SageMaker endpoint auto scaling questions: high-resolution metrics for faster detection and the shorter cooldown for faster follow-up scaling. Eliminate options that pair a 10-second interval with a standard-resolution metric or a longer 600-second cooldown.
Frequently Asked Questions
Why is the 600-second scale-in cooldown worse than the default 300 seconds?
A longer cooldown keeps the endpoint waiting longer before another scaling action, which reduces responsiveness when demand is volatile; 300 seconds supports faster adjustment.
Does SageMaker auto scaling support 10-second high-resolution metrics?
Yes. SageMaker endpoint target tracking can use high-resolution metrics at 10-second intervals, which detect traffic spikes sooner than standard one-minute metrics.
Related Analysis
Practice All MLA-C01 Questions
Access 115 questions with complete answers and detailed explanations.
View Full MLA-C01 Practice Test →