AWS Certified Solutions Architect – ProfessionalDesign for New SolutionsMedium

A manufacturing company wants to implement predictive maintenance for its industrial equipment. Sensors on the equipment generate telemetry data continuously, which needs to be ingested, processed in real-time to detect anomalies, and then stored for historical analysis and machine learning model training. The solution must be highly scalable, cost-effective, and provide insights with minimal operational overhead. Which architecture should a Solutions Architect recommend?

  1. AAWS IoT Greengrass for edge processing, Amazon SQS for queuing, Amazon EC2 for historical analysis, and Amazon S3 Intelligent-Tiering for storage.
  2. BAWS IoT Core for ingestion, Amazon Kinesis Data Streams for real-time processing, Amazon EC2 for anomaly detection, and Amazon RDS for storage.
  3. CAWS IoT Core for ingestion, AWS Lambda for real-time processing and anomaly detection, Amazon S3 for raw data storage, and Amazon Redshift for historical analysis.
  4. DAmazon Kinesis Data Firehose for ingestion, Amazon EMR for batch processing, Amazon DynamoDB for time-series data, and Amazon QuickSight for visualization.
Show answer & explanation

Correct answer: C. AWS IoT Core for ingestion, AWS Lambda for real-time processing and anomaly detection, Amazon S3 for raw data storage, and Amazon Redshift for historical analysis.

This architecture provides a serverless, scalable, and cost-effective solution. AWS IoT Core handles device connectivity and ingestion. AWS Lambda processes data in real-time for anomaly detection without managing servers. Amazon S3 offers highly scalable and cost-effective storage for raw data. Amazon Redshift is ideal for historical analysis and aggregates, supporting BI and ML training.

Why the other options are wrong

  • A. IoT Greengrass is for edge processing, which is useful but doesn't replace the cloud architecture for central analysis. SQS is a message queue, not ideal for real-time stream processing and high-volume ingestion. EC2 for historical analysis adds operational overhead compared to Redshift. S3 Intelligent-Tiering is a storage class, not the primary storage solution for raw data lake.
  • B. Kinesis Data Streams is suitable for real-time, but EC2 for anomaly detection adds operational overhead. RDS is not ideal for petabyte-scale time-series data and historical analysis due to cost and scalability limitations compared to S3 and Redshift.
  • D. Kinesis Data Firehose is good for ingestion but primarily for delivery to destinations, not real-time processing itself. EMR is for batch processing, not real-time anomaly detection. DynamoDB can store time-series but S3/Redshift combination is more common for raw storage and historical analysis at scale. QuickSight is for visualization, not the core processing/storage.

IoT Predictive Maintenance Architecture

An AWS architecture for ingesting, processing, and analyzing IoT sensor data in real-time to predict equipment failures and optimize maintenance schedules.

  • AWS IoT Core for secure device connectivity and data ingestion.
  • AWS Lambda for serverless real-time data processing and anomaly detection.
  • Amazon S3 for cost-effective, scalable raw data storage (data lake).
  • Amazon Redshift for historical analysis, aggregation, and ML model training.

Memory trick: Connect with IoT, process with Lambda, store in S3, analyze with Redshift.

More Design for New Solutions questions