AWS Certified Data Analytics – Specialty — Question 71

A company has an application that ingests streaming data. The company needs to analyze this stream over a 5-minute timeframe to evaluate the stream for anomalies with Random Cut Forest (RCF) and summarize the current count of status codes. The source and summarized data should be persisted for future use.
Which approach would enable the desired outcome while keeping data persistence costs low?

Answer options

• A. Ingest the data stream with Amazon Kinesis Data Streams. Have an AWS Lambda consumer evaluate the stream, collect the number status codes, and evaluate the data against a previously trained RCF model. Persist the source and results as a time series to Amazon DynamoDB.
• B. Ingest the data stream with Amazon Kinesis Data Streams. Have a Kinesis Data Analytics application evaluate the stream over a 5-minute window using the RCF function and summarize the count of status codes. Persist the source and results to Amazon S3 through output delivery to Kinesis Data Firehouse.
• C. Ingest the data stream with Amazon Kinesis Data Firehose with a delivery frequency of 1 minute or 1 MB in Amazon S3. Ensure Amazon S3 triggers an event to invoke an AWS Lambda consumer that evaluates the batch data, collects the number status codes, and evaluates the data against a previously trained RCF model. Persist the source and results as a time series to Amazon DynamoDB.
• D. Ingest the data stream with Amazon Kinesis Data Firehose with a delivery frequency of 5 minutes or 1 MB into Amazon S3. Have a Kinesis Data Analytics application evaluate the stream over a 1-minute window using the RCF function and summarize the count of status codes. Persist the results to Amazon S3 through a Kinesis Data Analytics output to an AWS Lambda integration.

Correct answer: B

Explanation

The correct answer is B, as it directly uses Kinesis Data Analytics to process the data stream in a 5-minute window, which aligns with the requirement for anomaly detection and status code summarization. Options A and C involve AWS Lambda for processing, which may not be as efficient for continuous streaming data analysis compared to Kinesis Data Analytics. Option D incorrectly uses a 1-minute window for evaluation, which does not meet the specified 5-minute timeframe.