In 2017 re:Invent IoT was one of the main topics as IoT Core got supporting services around it. The services released last year could be seen as basic components to create production ready industrial internet of things solutions in a secure and scalable manner. This year the released services revolved around helping build business logic around IoT data. Three major releases were SiteWise, Things Graph and Events, which all are in preview.
To help connect industrial sites to AWS, a service called SiteWise was released. It allows one to connect industrial network with gateway or Snowball Edge using OPC-UA protocol to AWS. It is non-invasive to factory network, subscribing data and sending them to IoT Core for further analysis. On top of connectivity, it allows you to model your assets and equipment in the cloud and add context to them. From here you can create metrics and display them on dashboards.
To ease the integration at the sensor level, AWS Things Graph offers a layer between Greengrass gateway and sensors. Layer is done by describing your sensors and services as models. These models have actions, states and events that are exposed as API written in GraphQL. To interact with these models one can create data flows in visual designer that act as step functions for IoT data. All this logic can be pushed then to Greengrass and run the logic at the edge.
AWS IoT Events on the other hand helps you detect events in the data stream and create actions and triggers on them. Detection is done by custom authored jsons, that model your equipments and processes and apply mathematical functions to process the data. Once a event is detected in can then be consumed by multiple AWS services to trigger actions. With AWS IoT Events one can build business logic for the IoT devices with serverless fashion.
There were two other releases that are not directly under IoT offering, but effect field of IoT data handling in form of time series data handling. First one is time series database service called AWS Timestream, which is released in preview. Time series databases in general are optimized for timestamped data. Time series data has specific characteristics and by utilizing them databases are able to outperform relational databases by thousand fold with time series data. AWS built Timstream from scratch and it is serverless. One could have hosted time series databases on their own before, but as a service this eases the management by large. AWS Timestream can be seen as last missing piece of the basic AWS IoT services.
Second service is called Amazon Forecast and it is also released in preview. AWS Forecast is service that consumes time series data and predicts future points. One can define the domain and feed data to the service which can automatically identify best fit model to create forecasts. Forecast comes with Amazon own developed models as well as well known models. This continues the long list of machine learning services AWS has released recently.
IoT related releases in 2018 re:Invent reflect the IoT market in general. Basic building blocks are now in place, technology is not an issue and focus is on building business solutions around them. Business solutions seem to revolve around predictive and preventative maintenance and efficiency improvement and by no surprise the main segment seems to be industrial applications. IoT is not anymore about inventing, it is about utilizing the services for business.