2.5Analytics, ML, IoT, Media, and Integration Services

These are best separated by intent (ingest/transform/search/visualize/infer/integrate). Each service is organized by definition, role, and when to choose.

2.5.1Analytics

• Amazon Athena: serverless analytics to query S3 data directly with standard SQL. Choose it for ad-hoc analysis of logs/data lakes with no ETL.
• Amazon EMR: managed clusters to run big-data frameworks like Spark/Hadoop. Choose it for large-scale distributed ETL or ML preprocessing.
• Amazon OpenSearch Service: managed search engine for full-text search, log analytics, and observability. Choose it for log search and dashboards (formerly Elasticsearch/Kibana).
• Amazon Kinesis Data Streams: high-throughput real-time stream ingestion. Choose it for streaming needing ordering, multiple consumers, and replay.
• Amazon Data Firehose: fully managed delivery that auto-loads/transforms streams into S3/Redshift/OpenSearch. Choose it to land streams into stores with no operations.
• Amazon MSK: managed Apache Kafka. Choose it for streaming where you want to keep existing Kafka assets/ecosystem.
• Amazon Managed Service for Apache Flink: real-time stateful stream processing (aggregation/analytics) with Apache Flink. Choose it to aggregate or detect anomalies on data as it arrives.
• AWS Lake Formation: builds data lakes and centrally governs access. Choose it to manage column/row-level permissions on an S3 data lake centrally.
• Amazon QuickSight: serverless BI dashboards/visualization. Choose it to share analytics as charts for business stakeholders.
• AWS Data Exchange: a marketplace to subscribe to and obtain third-party datasets. Choose it to bring external commercial data into analytics pipelines.
• Amazon AppFlow: no-code data transfer between SaaS (e.g., Salesforce) and AWS. Choose it to sync SaaS data into S3/Redshift on a schedule.

2.5.2Machine learning

• Amazon SageMaker AI: fully managed ML platform to build, train, and deploy models end to end. Choose it to train custom models and host inference endpoints.
• Amazon Rekognition: image/video analysis (objects, faces, moderation). Choose it to add image recognition without training.
• Amazon Comprehend: NLP (sentiment/entities/key phrases). Choose it to extract meaning from text.
• Amazon Textract: extracts text/tables/forms from documents (OCR+). Choose it to turn scanned forms into structured data.
• Amazon Transcribe: speech-to-text (ASR). Choose it to transcribe calls/meetings.
• Amazon Translate: neural machine translation. Choose it to translate content in real time or batch across languages.
• Amazon Polly: text-to-speech. Choose it to generate narration or voice responses.
• Amazon Kendra: ML-powered enterprise search (natural-language queries). Choose it to search internal documents in natural language.
• Amazon Personalize: ML service for real-time recommendations. Choose it for per-user recommendations/personalization.
• Amazon Fraud Detector: ML specialized in online fraud detection. Choose it to score payment/sign-up fraud with rules + models.

2.5.3IoT

• AWS IoT Core: the IoT core to securely connect devices and exchange messages (MQTT, etc.). Choose it as the platform to connect many devices to the cloud.
• AWS IoT Greengrass: runs local compute/inference/sync at the edge (device side). Choose it for on-site processing where connectivity is intermittent or latency matters.
• AWS IoT SiteWise: collects, models, and visualizes industrial equipment data. Choose it to aggregate factory OEE/metrics.
• AWS IoT Device Management: registers, organizes, and bulk-updates (OTA) many devices. Choose it to operate device fleets efficiently.
• AWS IoT Device Defender: security auditing and anomaly detection for devices. Choose it to monitor configuration drift or suspicious behavior across a fleet.
• AWS IoT Events: detects state from device events and triggers responses. Choose it to infer conditions (e.g., failure states) from multiple sensors.
• AWS IoT Things Graph: connects devices and cloud services via workflows (visual orchestration). Note: discontinued in 2024 (out of current scope).
• AWS IoT 1-Click: triggers Lambda from simple button-type devices. Choose it for one-touch actions from preconfigured devices. Note: discontinued in 2024.

2.5.4Media, integration, and others

• Amazon Kinesis Video Streams: ingest, store, and play back video/audio streams. Choose it to stream camera feeds for ML analysis or playback.
• AWS Elemental MediaConvert: transcodes media files (format/resolution). Choose it to convert uploaded video into delivery formats (note: the older Amazon Elastic Transcoder is closed to new use and being retired).
• AWS AppSync: managed GraphQL API (real-time/offline sync). Choose it to combine multiple data sources behind one endpoint for mobile/web.
• Amazon MQ: managed message broker (ActiveMQ/RabbitMQ). Choose it when existing apps use standard protocols like JMS/AMQP (use SQS/SNS for new decoupling).
• Amazon Managed Blockchain: managed blockchain networks (Hyperledger Fabric/Ethereum). Choose it to share a tamper-resistant ledger among multiple parties.
• Amazon Simple Email Service (SES): send/receive bulk transactional/marketing email. Choose it to send app notification email with high deliverability.

2.5.5Section summary

• Pick analytics by intent (Athena/EMR/OpenSearch/Kinesis/Firehose/MSK/Flink/Lake Formation/QuickSight)
• ML is task-specific APIs (Rekognition/Comprehend/Textract/Transcribe/Polly/Translate/Kendra/Personalize/Fraud Detector) plus general-purpose SageMaker AI