AWS Solutions Architect Associate Practice Test

What is the best way to ensure that an application scales to match demand when performance degradation occurs at high thread counts?

1. Launch two to six additional instances outside of the AutoScaling group to handle the additional load.

2. Populate a custom CloudWatch metric for concurrent sessions and initiate scaling actions based on that metric instead of on CPU use.

3. Empirically determine the expected CPU use for 200 concurrent sessions and adjust the CloudWatch alarm threshold accordingly.

4. Add more threads to the application to improve CPU utilization.

The correct answer is: Populate a custom CloudWatch metric for concurrent sessions and initiate scaling actions based on that metric instead of on CPU use.

Choosing to populate a custom CloudWatch metric for concurrent sessions and initiating scaling actions based on that metric effectively focuses on the actual performance parameters that are critical for the application. When performance degradation occurs at high thread counts, simply monitoring CPU usage may not provide an accurate representation of the application's state or its ability to handle incoming requests. By creating a custom CloudWatch metric that directly reflects the number of concurrent sessions, you can better monitor the application's operational demand. This approach allows for more precise scaling actions, ensuring that more resources are allocated in response to high traffic before performance begins to degrade significantly. It aligns the scaling strategies with the actual usage pattern rather than relying on potentially misleading metrics like CPU utilization, which may not correlate directly with performance issues under certain workload conditions. Utilizing concurrent sessions as a scaling trigger can optimize resource allocation, thereby improving user experience and application availability under peak loads. This strategy is especially beneficial in scenarios where the application is I/O bound or has varying performance characteristics that are not directly tied to CPU usage. The other approaches could lead to either inefficient resource management or not addressing the root cause of the performance issues during high demands. For instance, launching additional instances outside of an AutoScaling group may create unnecessary complexity and fail to automate scaling efficiently. Determ