MLA-C01 Question 271
Single answerYou are tasked with deploying a machine learning model on AWS SageMaker. The model needs to handle real-time low-latency predictions for a mobile application, but the traffic is unpredictable, ranging from zero to thousands of requests per second. Which deployment option should you choose?
- A
Real-time endpoint with auto-scaling
- B
Serverless inference endpoint
- C
Batch transform
- D
Asynchronous inference endpoint
Show answer and explanation
Correct answer: B
Explanation
Serverless inference endpoints in SageMaker are specifically designed for unpredictable traffic patterns. They scale automatically based on demand, ensuring low latency for real-time applications without incurring costs during idle periods. This makes them an ideal choice for applications with variable traffic, such as the mobile app described in the scenario.
- A. Incorrect.
Real-time endpoints with auto-scaling can handle varying traffic, but they require pre-provisioned instances, which may lead to costs even when there is no traffic.
- B. Correct.
Serverless inference endpoints are suitable for unpredictable workloads because they automatically scale to zero when idle and scale up to handle sudden traffic spikes, making them cost-efficient and ideal for this use case.
- C. Incorrect.
Batch transform is designed for offline, large-scale batch processing and is not suitable for real-time low-latency predictions.
- D. Incorrect.
Asynchronous inference endpoints are designed for processing requests that do not require immediate responses and are not ideal for low-latency real-time applications.