AWS Certified Machine Learning Engineer Associate · Free Practice Question Hard
Question 16
A telecommunications company receives real-time network performance data streams from thousands of devices across its infrastructure. The data streams consist of thousands of JSON records per second, detailing metrics like latency, packet loss, and jitter. The company needs to implement a scalable solution on AWS to monitor these metrics and detect anomalies in real time with the least operational overhead.
Which solution will meet these requirements?
-
A
Use AWS Lambda to process the network metrics streams in real time and write a custom function to detect anomalies directly from the data
-
B
Use Amazon Kinesis Data Streams to ingest the data, process it with Apache Flink to analyze the streams in real time, and apply the RANDOM_CUT_FOREST algorithm to detect anomalies
-
C
Ingest the data using Amazon Kinesis Data Streams, analyze the network metrics using Amazon Comprehend, and identify anomalies with a SageMaker model
-
D
Use Amazon Kinesis Firehose for real-time streaming of the network metrics data into an Amazon S3 bucket, then periodically run an AWS Glue ETL job to detect anomalies using a custom Python script
Reveal correct answer
Correct answer: B
Explanation
Correct option:
Use Amazon Kinesis Data Streams to ingest the data, process it with Apache Flink to analyze the streams in real time, and apply the RANDOM_CUT_FOREST algorithm to detect anomalies
This solution uses AWS services designed for real-time processing and anomaly detection:
Amazon Kinesis Data Streams - Provides a scalable, low-latency service for ingesting high-volume, real-time network metrics.
Apache Flink on Kinesis Data Analytics - Processes the data streams in real time and offers powerful frameworks for data analysis.
RANDOM_CUT_FOREST algorithm - A built-in anomaly detection algorithm within Apache Flink that minimizes the need for custom code or external models.
This combination handles high-throughput data streams and detects anomalies efficiently, with minimal operational overhead.
Incorrect options:
Use Amazon Kinesis Firehose for real-time streaming of the network metrics data into an Amazon S3 bucket, then periodically run an AWS Glue ETL job to detect anomalies using a custom Python script - Kinesis Firehose delivers data in real-time to destinations like Amazon S3 but using AWS Glue for periodic jobs introduces latency and is unsuitable for real-time anomaly detection.
Ingest the data using Amazon Kinesis Data Streams, analyze the network metrics using Amazon Comprehend, and identify anomalies with a SageMaker model - Amazon Comprehend is a natural language processing tool and cannot analyze numerical or structured data like network metrics. Additionally, SageMaker for real-time analysis adds unnecessary complexity compared to Apache Flink’s built-in capabilities.
Use AWS Lambda to process the network metrics streams in real time and write a custom function to detect anomalies directly from the data - While Lambda can handle real-time data, it is not optimized for high-throughput or resource-intensive computations like anomaly detection. Scaling Lambda for such tasks adds operational complexity.
References:
https://aws.amazon.com/managed-service-apache-flink/
https://docs.aws.amazon.com/firehose/latest/dev/what-is-this-service.html
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
