MLA-C01 Question 402
Select 4You are managing a machine learning model deployment on Amazon SageMaker. The workload involves real-time inference with fluctuating traffic patterns. Which key performance metrics should you monitor to ensure the system meets business requirements while maintaining efficiency and reliability?
- A
Model endpoint availability
- B
CPU and GPU utilization of the deployed instances
- C
Number of inference requests processed per second (throughput)
- D
Data preprocessing accuracy
- E
Scalability of the endpoint to handle increased traffic
Show answer and explanation
Correct answers: A, B, C, E
Explanation
Key performance metrics for ML infrastructure include endpoint availability, resource utilization (CPU/GPU), throughput, and scalability. These metrics help ensure that the deployed model performs efficiently, scales with demand, and meets reliability and performance expectations. Data preprocessing accuracy, while important in model development, is not a direct performance metric for deployed ML infrastructure.
- A. Correct.
Monitoring model endpoint availability ensures the service is accessible to end users and meets uptime requirements, which is a critical factor in reliability.
- B. Correct.
CPU and GPU utilization helps in assessing the efficiency of resource usage and identifying underutilized or overburdened instances.
- C. Correct.
Throughput is essential to measure how effectively the system processes inference requests, especially under varying traffic loads.
- D. Incorrect.
Data preprocessing accuracy is related to data preparation and not directly tied to ML infrastructure performance metrics.
- E. Correct.
Scalability is crucial for ensuring that the system can dynamically handle increased or decreased traffic without performance degradation.