After training a time-series forecasting model in SageMaker, a Specialist needs to run load tests on the endpoint and monitor latency, memory, and CPU during the test. Which approach provides a consolidated view of these metrics during load testing?

Creating an Amazon CloudWatch dashboard provides a consolidated, real-time view of latency, memory usage, and CPU metrics. SageMaker automatically publishes these metrics to CloudWatch, making them readily available for dashboarding without additional configuration. This approach is efficient for monitoring during load testing because CloudWatch is designed for real-time operational data. Writing SageMaker logs to S3 and using Athena/QuickSight is better suited for historical analysis of log data, not real-time metric monitoring. Sending CloudWatch Logs to Amazon Elasticsearch Service (Amazon ES) and using Kibana is effective for detailed log analysis and searching, but CloudWatch dashboards are more direct and performant for visualizing pre-defined metrics. Exporting SageMaker-generated CloudWatch Logs to Amazon ES is redundant as CloudWatch already contains the necessary metrics.