Enterprise cloud providers, such as AWS, Google, and Microsoft, are targeting the data-rich industrial IoT edge as a primary means of delivering analytics-driven business improvement. The industrial IoT edge is where most of the data necessary for analytics resides, plus there is tremendous value in providing preprocessed analytics output to the cloud – if not executing closed loop ML at the edge itself, even in real-time.
Cloud providers are positioning edge offerings to their IT customers based on a common edge development environment and availability of large developer and support ecosystems relative to dedicated automation environments. Extending their reach to the edge further allows cloud providers to leverage their strengths in data architecture, integration, and security. One of the most compelling developments to watch in this area is the potential impact of microcontroller hardware offerings announced by the enterprise cloud players in support of this migration.
Industrial IoT Edge Offerings from Enterprise Cloud Providers Include Microcontrollers
All three major enterprise cloud providers have announced their intent to provide hardware to enable data-intensive edge applications, such as video analytics for surveillance and inspection. Third party chips, such as NVIDIA, can also typically be employed. It is easy to see how this capability could extend to displacement of non-deterministic logic currently performed by dedicated industrial automation equipment.
In late 2018, AWS announced it has designed its own processors, called AWS Inferentia, for inference in edge-based machine learning applications. This microcontroller can be used with the company’s AWS FreeRTOS to extend connection to AWS IoT Greengrass Core devices all the way down to industrial endpoints.
Google’s Cloud IoT edge is an emerging component of an integrated stack that includes the Linux OS and Google’s Edge TPU microcontroller. The solution is intended to enable IoT edge devices to process and analyze images, videos, gestures, acoustics, and motion locally rather than send raw data to the cloud. The EdgeML component is designed to significantly increase the processing power and versatility of edge devices, while also reducing response latency, by enabling local inference of TensorFlow Lite ML models pre-trained in the cloud and executed on a CPU, GPU, or Google’s own Edge TPU in a gateway or endpoint device.
Microsoft Azure Sphere is a new solution for creating highly-secured, Internet-connected microcontroller (MCU) devices. Azure Sphere includes three components that work together to protect and power devices at the intelligent edge: MCUs, an OS, and Sphere Security Service. Azure Sphere MCUs combine real-time and application processors with built-in Microsoft security technology and connectivity.
Material for this blog was generated from ARC’s new global market research report on Industrial IoT Edge Software Platforms. For more information on this and other available ARC market research on the industrial IoT edge, see our dedicated IIoT edge research webpage. You can also join the conversation about this exciting topic on ARC’s blog sites or LinkedIn Groups: IIoT and IIoT Network Edge.
About ARC Advisory Group (www.arcweb.com): Founded in 1986, ARC Advisory Group is a Boston based leading technology research and advisory firm for industry and infrastructure.
For further information or to provide feedback on this article, please contact RPaira@arcweb.com
About the Author:
Chantal’s focus areas include the Industrial Internet of Things (IoT), industrial Ethernet switches and devices, wireless networks, device networks, and intelligent train control and rail signaling. She also administers the ARC Industrial Internet of Things group on LinkedIn. Chantal has been with ARC since 1990 and has conducted numerous industry-leading research activities in areas including: Connected Device Management Platforms for Industrial IoT, Industrial Ethernet Devices, Industrial Ethernet Switches and Infrastructure, Industrial Wireless (process and discrete industries), Industrial Device Networks, Intelligent Train Control Systems.