‹ All AI-300 questions

AI-300 Practice Question 63: Online Endpoint Autoscale

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 7 · question 3 of 10

Implement machine learning model lifecycle and operationsmedium

A managed online endpoint runs a deployment with a fixed instance count of 2. CPU usage spikes every weekday morning and latency rises. The team wants more instances to be added automatically when average CPU utilization exceeds 70 percent, and removed afterwards, without redeploying the model. What should they configure?

Select one answer
Show answer and explanation

Answer:

  • A. Azure Monitor autoscale with a CPU utilization rule

Managed online endpoints integrate with Azure Monitor autoscale. A scale-out rule on the CpuUtilizationPercentage metric adds instances when the average passes the threshold, and a matching scale-in rule removes them later. The instance count follows the load with no redeployment (option A).

- B is wrong because managed online endpoints run on compute managed by the service. They do not scale by using a compute cluster that you create. - C is wrong because this is not automatic and pays for 10 instances all day, including the hours when the load is low. - D is wrong because data collection stores request payloads for monitoring. It does not change the number of instances.

Reference: Autoscale an online endpoint

Practice Test 7 ›

Study guides for AI-300