Organised data archival and analysis for prediction and improving productivity is a challenge for many industries. The architecture referenced can be used for live data ingestion and breakdown into reports with machine-learning capabilities. For a traditional customer premises system when migrating into this massive scale data storage and processing this architecture requires some modifications which will also be explored as case studies.
ETL Edge though depicted with an AWS Snowball Edge device, this could even be swapped with IOT sensors and scanners polled in through a Raspberry PI which has the AWS IOT Core libraries or ulitizes the message broking MQTT to dynamically send telemetry data, to be ingested into the warehouse and to be processed by machine learning process in Amazon SageMaker. This is just a reference architecture which requires further polishing to adapt it to any real-world situations. We can explore some use cases.
Transport Service
Yes this is a possibility to make it easy for the transport service to optimize route management, predict maintenance proposals and proactively take decisions. For a start the AWS Snowball Edge can be swapped for data enabled ticketing system which could push the inline ticketing information between segments. Public user search and feedback can be pushed from the ecom enabled web portal. Trip adjustments can be predicted with enough data and proper Machine Learning Models.
Restaurant Chain
Edge device would be a mobile or web application which integrates in house dining orders, delivery orders, billing, vendor system for inventory management, Amazon alexa support could be integrated with the whole data converging into the customized solution which can help in periodic and seasonal shift in cusine categories and help alerting vendors with requirements in time to take the human errors out, and let us concentrate on the business.
Hospital or Healthcare
Edge device integration can be done for all sorts of digital imaging systems, monitoring systems, data processing and analytical equipment. Machine Learning and Artificial Intelligence could assist in much faster diagnosis. I do agree that nothing can replace an experienced human who knows his job, but such solutions will provide as a suggestion only. Medical monitoring or imaging could start ingesting petabytes of data per day when you consider the proprietory image formats with metadata permitting to zoom in and measure.
Chain of Supermarkets
Edge devices could be pos machines, mobile applications of Vendors, store cc tv, exit rf scanners in each sections. By utilizing proper video/image analysis in real time, could mitigate many complications, and assist in prediction inventory flexing
Pretty sure that this is an incomplete stub, but will update in the near future.