AWS re:Invent has now been arranged for five years in a row and the event has a reputation for introducing brand new AWS services or additions to the existing ones. The main themes for the 2017 re:Invent were data driven applications, machine learning and IoT. Two of the latter ones were very tightly connected to everything new and the culmination could be seen as the release of the DeepLens, the world’s first deep learning enabled video camera for developers. In the context of IoT, here are the main releases of the 2017 re:Invent event.
Data, IoT and ML
IoT is big! Gartner has forecasted that 8.4 billion connected things will be in use in 2017. The market is estimated to be around 1.2 trillion dollars. IoT devices, as we see them today, are still simple sensor devices sending simple sensor data. In 2016 there were estimated 20 billion microcontroller and only some 0.5 billion cpu powered IoT devices. This means that the relation is around 4:1 in favor of microcontroller chips.
The existing fleet of microcontroller chips could be further developed to the use of IoT, but they require updates in the cpu level to be able to connect to a cloud. To solve this problem AWS announced freeRTOS which is created to help microcontroller development. They have even got the designer of freeRTOS on board, which is a serious statement on its own already.
Of course using FreeRTOS is a deep dive and for a simple IoT cases it would be too heavy. For a popular demand AWS also introduced 1-click IoT devices which, to be honest, really seems just a relaunch of IoT button. Regardless, with this single button one can leverage IoT very easily and fast. Bad news is that the connectivity is provided with AT&T by default, making the product mainly suitable just for the US market.
#1 Greengrass upgrade
Typically, sensors require a gateway to connect to the cloud or if the device is CPU powered with connection, it needs a software to run. With the release of the greengrass last year one could run lambda code on the device.
New features introduced to the Greengrass includes over the air updates to the device and the access to the local resources from the lambda code, making the sensors more directly available.
On top of these improvement there is ability to run ML on the edge. With the announcement of the SageMaker, a ML service allows now users to build, train and deploy models. These models can be run on the Greengrass, pushing the analytics to the edge.
#2 Fleet management is available now
Previously, the ability to manage devices on at the AWS IoT Core was very limited. Basicly, you had a device registry, which was simple way to organize devices, but provided basicly just a list of devices.
With the announcement of AWS Device Management one can now securely onboard, organize, monitor, and remotely manage one’s IoT devices.
Device Management has a big impact on how the Greengrass can be used, deployed and managed over the lifecycle of devices. It also makes the AWS IoT platform ready for industrial scale deployments. It will be interesting to see, what kind of devices and sensors will come to the market leveraging the Greengrass and FreeRTOS sensors Device Management is generally available now.
#3 Better security
Security is important – so important that CTO Dr. Vogel spend great time to talk about it in his keynote.
New AWS IoT Device Defender announcement came with services to audit device configurations, monitor device behaviour, identify anomalies and create alerts to protect the IoT fleet.
All of these security features are vital when you are scaling your fleet of IoT devices to hundreds, thousands or even to millions of devices. The audit service monitors policies and checks that security settings are in their place and will supports the custom audit rules. Identifying anomalies at high scale fleet would be impossible without anomaly detection. Service uses ML to identify odd behaviour and allows triggering alarms on these events.
If one already has a fleet of devices in one place, the bad news is that you need to wait until 2018 to get Defender to protect your fleet.
#4 Intelligence
Machine learning, and deep learning especially, were talked about at the central stage at the re:Invent with big announcements related to it. Before re:Invent AWS announced a great feature to the IoT rules engine, allowing lambda calls directly from the SQL. With the lambda, you can enrich the incoming data with for example running ML model or extract data from existing data sources.
Intelligence side the major announcement was AWS IoT Analytics.
IoT Analytics offers services to process the noisy data from sensors through pipelines, store the data to time series database, run adhoc and in depth queries and create templated reports on Quicksite. One can use the IoT Analytics pipelines and processes to IoT data as well as data coming through Amazon Kinesis. This will simplify analysis of the IoT dataand gives ability to react more quickly to the incoming data. IoT Analytics is in preview mode.
IoT is here to stay
With the announcements mentioned above, AWS is now offering all parts of the IoT pipeline, all the way from the sensors to gateways to cloud and finally to machine learning. They also make the AWS IoT platform ready to be scaled in managed and secured way. Keep in mind, that AWS builds services on customer demand. I’d say that the IoT is not anymore somewhere in thefuture. It’s here already.
Antti Loukiala works at Solita as a Data Engineer in Agile Data IoT specialised team. He believes that IoT is not only changing the way companies use and gather data, but how companies do business. Antti thinks that this transition is what makes IoT extremely interesting to work with.
