Access the latest quantum technology

Quantum technology in Bristol and bath - find out more about how you can access the commercialisation of quantum technology for sensing and security

Friday, December 01, 2017

Amazon brings machine learning to the Edge

By Nick Flaherty

Amazon Web Services has announced six significant services and capabilities for connected devices at the edge that will have a significant impact on embedded systems design. 

AWS IoT 1-Click, AWS IoT Device Management, AWS IoT Device Defender, AWS IoT Analytics, Amazon FreeRTOS, and AWS Greengrass ML Inference are intended to simplify deployment and management of large fleets of devices, auditing and enforcement of consistent security policies, and analysis of IoT device data at scale. 

AWS Greengrass ML Inference allows machine learning models to be deployed directly to devices, where they can run machine learning inference to make decisions quickly, even when devices are not connected to the cloud. 

“The explosive growth in the number and diversity of connected devices has led to equally explosive growth in the number and scale of IoT applications. Today, many of the world’s largest IoT implementations run on AWS, and the next phase of IoT is all about scale as we’ll see customers exponentially expand their fleet of connected devices,” said Dirk Didascalou, VP IoT, AWS. “These new AWS IoT services will allow customers to simply and quickly operationalize, secure, and scale entire fleets of devices, and then act on the large volumes of data they generate with new analytics capabilities specifically designed for IoT. With Amazon FreeRTOS, we’re making it easy for customers to bring AWS IoT functionality to countless numbers of small, microcontroller-based devices. And, customers have also told us they want to execute machine learning models on the connected devices themselves, so we’re excited to deliver that with AWS Greengrass ML Inference.”

With AWS IoT 1-Click, enabling a device with an AWS Lambda database function is as easy as downloading the mobile app, registering and selecting an AWS IoT 1-Click enabled device, and – with a single click – associating an AWS Lambda function. AWS IoT 1-Click comes with pre-built AWS Lambda code for common actions like sending an SMS or email. Customers can also easily author and upload any other Lambda function.

AWS IoT Device Management and AWS IoT Device Defender simplify onboarding, managing, and securing fleets of IoT devices, while AWS IoT Analytics makes it easy to run sophisticated analytics on the data generated by devices.

AWS IoT Device Management (available today) makes it easy to securely onboard, organize, monitor, and remotely manage IoT devices at scale throughout their lifecycle—from initial setup, through software updates, to retirement. Getting started is easy; customers simply log into the AWS IoT Console to register devices, individually or in bulk, and then upload attributes, certificates, and access policies. Once devices are in service, AWS IoT Device Management allows customers to easily group and track devices, quickly find any device in near real-time, troubleshoot device functionality, remotely update device software, and remotely reboot, reset, patch, and restore devices to factory settings, reducing the cost and effort of managing large IoT device deployments.
AWS IoT Device Defender (coming in the first half of 2018) continuously audits security policies associated with devices to make sure that they aren’t deviating from security best practices, and alerting customers when non-compliant devices are detected. AWS IoT Device Defender also monitors the activities of fleets of devices, identifying abnormal behavior that might indicate a potential security issue. For example, a customer can use AWS IoT Device Defender to define which ports should be open on a device, where the device should connect from, and how much data the device should send or receive. AWS IoT Device Defender then monitors device traffic and alerts customers when anomalies are detected, like traffic from a device to an unknown IP address.
AWS IoT Analytics (available in preview) is a fully managed analytics service that cleans, processes, stores, and analyzes IoT device data at scale. Getting started is easy: customers simply identify the device data they wish to analyze, and they can optionally choose to enrich the device data with IoT-specific metadata, such as device type and location, by using the AWS IoT Device Registry and other public data sources. AWS IoT Analytics also has features for more sophisticated analytics, like statistical inference, enabling customers to understand the performance of devices, predict device failure, and perform time-series analysis. And, by using Amazon QuickSight in conjunction with IoT Analytics, it is easy for customers to surface insights in easy-to-build visualizations and dashboards.

Amazon FreeRTOS is an operating system that extends the rich functionality of AWS IoT to devices with very low computing power, such as lightbulbs, smoke detectors, and conveyor belts. 

“As we’ve seen the Arm-based microcontroller ecosystem grow over recent years, FreeRTOS has played a key role in enabling embedded developers,” said Rene Haas, EVP and President, IP Products Group (IPG) at ARM. “We are pleased to see AWS extend the FreeRTOS kernel with increased connectivity, while adding additional security features. Amazon FreeRTOS running on Arm-based processors is an important milestone toward improving hardware, software, and networking security for the industry.”

AWS Greengrass ML Inference is a new feature of AWS Greengrass that lets application developers add machine learning to their devices, without requiring special machine learning skills. IoT devices frequently collect and forward large quantities of data, which can be used to automate real-time decision making through machine learning. To do this, customers build, train, and run machine learning on their IoT data in the cloud. However, some applications are highly latency sensitive and require the ability to make decisions without relying on always-on network connectivity. 

With AWS Greengrass ML Inference, devices can run machine learning models to perform inference locally, get results, and then make smart decisions quickly, even when they’re not connected. Using Amazon SageMaker, or any machine learning framework, customers build and train their machine learning models in the cloud and then – with just a few clicks – use the AWS Greengrass console to transfer the models to devices they select.

No comments: