A company needs to implement a near-real-time fraud prevention feature for its ecommerce site. User and order details need to be delivered to an Amazon SageMaker endpoint to flag suspected fraud…

Question

Options

• ACreate an Amazon Managed Streaming for Kafka cluster and ingest the data for each order into a
• BCreate an Amazon Kinesis Data Streams stream and ingest the data for each order into the
• CCreate an Amazon Kinesis Data Firehose delivery stream and ingest the data for each order into
• DCreate an Amazon SNS topic and publish the data for each order to the topic. Subscribe the

• A
  75% (27)
• B
  3% (1)
• C
  14% (5)
• D
  8% (3)

Explanation

The critical constraint is the 1.5 MB payload size. Amazon Kinesis Data Streams has a hard 1 MB per-record limit, ruling out Option B. Amazon SNS has a 256 KB message size limit, ruling out Option D. Amazon Kinesis Data Firehose (Option C) buffers data before delivery, introducing latency that is incompatible with near-real-time requirements. Amazon MSK (Managed Streaming for Apache Kafka) can be configured to handle message sizes well above 1.5 MB (via the max.message.bytes configuration) and supports low-latency consumer applications that can invoke a SageMaker endpoint directly. MSK provides the combination of large payload support and minimal latency needed for this use case.

No community discussion yet for this question.