Azure IoT and Edge Computing announcements dominated the keynote at Microsoft Build, the annual developer conference. Microsoft is on track to deliver the promise of Intelligent Cloud and Intelligent Edge.
First outlined by Satya Nadella, CEO of Microsoft, at Build 2017, Intelligent Edge is a strategic bet for Microsoft. The company is taking the lead in designing the distributed architecture that makes Public Cloud more accessible. From leveraging containers to creating an app store equivalent of Edge Computing components, Microsoft is innovating for the edge.
Microsoft Azure IoT is a comprehensive enterprise IoT platform in the Public Cloud. Features such as Dynamic Provisioning Service (DPS) and Time Series Insights make it an industry-leading IoT PaaS. With Azure IoT Edge, Microsoft is pushing some of the most critical capabilities of its IoT PaaS to the edge.
As an advocate of decentralized and distributed Public Cloud, I am impressed with the architecture and the long-term strategy of Microsoft Intelligent Edge. Here are a few reasons why I believe Azure IoT Edge is built the right way.
Modular architecture based on containers
Azure IoT Edge platform is designed to run a set of components packaged and deployed as containers. Microsoft is shipping some of the core building blocks of Azure IoT such as Functions and Stream Analytics as container images. Each container deployed at the edge is called as a module. Developers can create custom container images, upload them to the registry, and then use the control plane to push those containers as modules to the edge. Similar to devices, modules also have a digital twin that’s accessible in the Public Cloud. Identical to UNIX Pipes, multiple modules can be chained together to perform data transformation, aggregation, filtering, and encryption.
Marketplace for edge modules
Microsoft is building an app store for Azure IoT Edge developers to publish custom modules. Businesses can buy and download these modules to their edge devices. For example, there may be modules for converting inbound telemetry sent via LoRA or Sigfox protocol, which customers can integrate with their deployments. The marketplace will become an authoritative repository of modules that can be linked together to achieve a specific business outcome.
Kubernetes as the control plane for edge modules
If there is more than one container to manage, Kubernetes becomes the de facto choice. Last year, Microsoft developed an open source bridge called Virtual Kubelet that connects provisioning and scheduling engines to Kubernetes. Microsoft’s container platform team has extended Virtual Kubelet to Azure IoT Hub to push edge modules through familiar Kubernetes orchestration engine. Since Azure IoT Hub is the device and module provisioning layer, Microsoft is able to seamlessly extend IoT Hub to Kubernetes. By using the primitives of Kubernetes, Azure IoT team made it easy for developers to package and deploy edge modules. This is one of the brilliant moves from Azure IoT and container teams.
Open sourcing Azure IoT Edge runtime
Microsoft realized that the edge ecosystem would grow exponentially in the coming years. There will be an explosion of new protocols, standards, privacy and security regulations, data formats, and device architectures. By open sourcing the Azure IoT Edge runtime, Microsoft is avoiding from becoming a potential bottleneck for innovation. Device gateway manufacturers, OEMs, telecom operators, silicon manufacturers can contribute to the open source edge platform. In the long-term, this becomes a key differentiator for Microsoft in the form of a vibrant ecosystem.
Bringing AI to the edge
Early in the development cycle, Microsoft realized AI would become the most significant driver for edge computing. The company is prioritizing on edge compatibility for all AI services. CustomVision.ai, Microsoft’s AutoML platform for computer vision can export trained models that can be run at the edge for inferencing. Microsoft is ensuring that Open Neural Network Exchange (ONNX) runtime is available on Azure IoT Edge to run models built with heterogenous deep learning frameworks.
Ecosystem engagement
The edge computing ecosystem is very dynamic and broad. From silicon manufacturers to industrial automation OEMs, everyone has a role to play. Microsoft is partnering with HPE, Advantech, and Moxa to provide secure edge hardware to enable secured IoT devices from chipset to the cloud. Microsoft and DJI, the drone company, are collaborating to develop commercial drone solutions, leveraging Azure IoT Edge and AI technologies for customers in industry verticals including agriculture, construction, and public safety. Microsoft has announced a joint effort with Qualcomm to create a vision AI developer kit that is similar to Amazon’s AI-powered camera, AWS DeepLens.
AWS Greengrass, Amazon’s Edge Computing platform is built to deliver similar capabilities as Azure IoT Edge. AWS Greengrass became generally available last year – much before the design of Azure IoT Edge. While both the platforms are designed to meet the key customer use cases, I feel that Azure IoT Edge has an edge over AWS Greengrass (pun intended).
The future of cloud lies at the edge. Microsoft’s investment in Intelligent Edge will help the company in driving the adoption of its Public Cloud.
