DOP-C02 Question 243
Select 2Your team has implemented an Amazon CloudWatch anomaly detection alarm to monitor the latency of an API hosted on AWS Lambda. During peak hours, you notice the alarm frequently triggers even though the latency remains within acceptable thresholds for your business. Which of the following actions should you take to address this issue while ensuring the alarm remains effective for detecting outliers?
- A
Adjust the anomaly detection model's sensitivity by configuring a higher 'upper bound' value.
- B
Increase the evaluation period for the anomaly detection alarm to reduce false positives.
- C
Manually modify the expected baseline metric values for the anomaly detection algorithm.
- D
Use a more stable metric, such as the average latency over a longer period, for the anomaly detection alarm.
- E
Enable auto-scaling for the Lambda function to reduce peak load and prevent the alarm from triggering.
Show answer and explanation
Correct answers: B, D
Explanation
To address false positives in anomaly detection alarms, you can either increase the evaluation period to account for more data points or use a more stable metric, such as average latency, to reduce the impact of short-term fluctuations. These actions ensure the alarm remains effective for detecting actual anomalies while minimizing unnecessary triggers.
- A. Incorrect.
Adjusting the upper bound value is not a feature of Amazon CloudWatch anomaly detection alarms. Sensitivity is managed automatically based on the metric data and cannot be manually tuned this way.
- B. Correct.
Increasing the evaluation period allows the alarm to evaluate more data points over time, reducing the likelihood of false positives during temporary spikes.
- C. Incorrect.
Manually modifying the baseline metric is not supported in Amazon CloudWatch anomaly detection alarms. The algorithm automatically learns and adjusts based on historical data.
- D. Correct.
Using a more stable metric like average latency over a longer period can smooth out short-term fluctuations and provide a more accurate metric for anomaly detection.
- E. Incorrect.
While auto-scaling can help reduce peak load, it does not directly address the configuration of the anomaly detection alarm. The issue in the scenario pertains to alarm configuration rather than application scaling.