Building The Intelligent Edge: Microsoft Azure Stack Edge

"Edge computing" and why we need it

Businesses increasingly rely on actionable insight and near-instant data inference to make key decisions. This kind of data-driven decision making is made possible by modern application architectures that can generate and process the rapidly increasing amounts of data being generated and gathered. Edge computing offers a distributed architecture to meet these needs, often operating on the "edge" of a client's network where the data processing is nearest to the data sources.

For example, consider a use case in manufacturing. A business may operate in dozens of different manufacturing facilities, each outfitted with thousands of IoT sensors that monitor manufacturing equipment and gather data on the status, health and operating parameters of this equipment.

There may also be specialized video cameras that monitor production all of the lines, all of which report on "out of spec" conditions. Sending this vast amount of data back to a centralized data center or public cloud environment over a network may be impossible, cost prohibitive or even risky if network reliability is a concern. Also, the increased delay in the output of the data processing may be unacceptable if near real-time operational and safety decisions are affected.

Edge computing presents as a solution to these challenges, providing:

• localized access to data sources;
• quicker creation of actionable insights;
• conservation of network resources; and
• less risk of adversely affecting near real-time operational and safety decisions.

Microsoft's vision of the "Intelligent Edge"

Microsoft Azure is available in 140 countries and operates in 54 data center regions worldwide. Even with Microsoft's investment in the "Intelligent Cloud," they've recognized a need to develop solutions that cater to the "Intelligent Edge." Microsoft doesn't see the Intelligent Cloud and Intelligent Edge as two discreet platforms, but rather as a ubiquitous computing platform that extends between public cloud offerings and on-premises solutions.

Azure Stack Edge

Microsoft's Azure Stack Edge is part of a family of Intelligent Edge solutions, and is a cloud-managed, AI-enabled edge computing device with data storage and transfer capabilities into Microsoft Azure. Azure Stack Edge was released to the public in the spring of 2019 and takes the form of a cloud-connected hardware appliance that can be installed locally on a business's network.

The Azure Stack Edge solution is comprised of three items:

1. The Azure Stack Edge Physical Appliance - The appliance is a 1RU node rack system that consists of 20 CPU cores and an Intel Arria 10 field-programmable gate array (FPGA) chipset.
2. An Azure Stack Edge Resource - This is created in the Azure portal, and allows businesses to manage the Azure Stack Edge appliance from a web interface.
3. The Azure Stack Edge Local UI - Azure Stack Edge provides a local UI for the appliance to run diagnostics, shut down/restart the device, view logging information, and can be used to interact with Microsoft support for service requests.

At a high-level, Azure Stack Edge seeks to provide the following key capabilities:

• Network Storage Gateway - Azure Stack Edge offers the ability to perform data transport into Azure while also retaining local access to files. For example, a business could expose an SMB or NFS share locally on the network that "synchronizes" data with the Azure public cloud. This capability also addresses a challenge businesses may have: getting data into the Azure public cloud. It's important to note here that Azure Stack Edge isn't a unique solution that can provide data transfer capabilities into the Azure public cloud – this can also be provided by the Azure Data Box, Data Box Disk, Data Box Heavy (all offline) or Data Box Gateway virtual appliance (online).
• Edge Compute/Data Pre-Processing - The Azure Stack Edge can run containerized applications managed through Microsoft's Azure IoT Edge/IoT Hub service. For example, one can deploy a Linux container solution locally on the Azure Stack Edge that provides data pre-processing for IoT sensors deployed on the same physical network, effectively "scrubbing" this data before it is sent to the Azure public cloud. Pre-processing can also be used to aggregate data, modify data or analyze and react to IoT events. Also announced recently was coming support for Kubernetes cluster and virtual machine deployments.
• Azure Machine Learning Powered by Intel Arria FPGA - This is specifically used in scenarios where the pre-processing of data would leverage Azure Machine Learning. This capability is also provided in the Azure public cloud, but physics/speed of light limitations might make it necessary to perform inferencing on data closer to the source of that data. Using Azure Stack Edge, businesses can run machine learning models built in the Azure cloud directly on localized hardware, providing quick results that can be acted upon before that data is sent to the cloud.
• Cloud Managed - All management of the Azure Stack Edge is performed through the Azure public cloud portal. This addresses the process of acquiring the Azure Stack Edge and performing routine maintenance on a (potentially) large installed base of Azure Stack Edge appliances.

Example use cases

Let's go back to the manufacturing example above and talk about a potential business need regarding video surveillance. Perhaps this manufacturer was recently fined for safety violations and is seeking ways to address this. They might decide to deploy video surveillance cameras to identify staff members not wearing appropriate safety protection. Leveraging the Azure Stack Edge solution, they can ingest these video feeds and run Azure Cognitive Services Vision models locally within the facility to identify and flag out of compliance conditions in real-time. There is also no need to send large amounts of video data over the network.

Another use case in manufacturing would involve detecting out of spec conditions through video feeds installed on a production line. Similar to the safety example, the manufacturer could leverage machine learning models locally on cloud-connected and cloud-managed infrastructure installed within the business's facility. This would allow for the means to identify out of spec conditions quickly, and without the need to transfer large amounts of data offsite.

One final use case could address a very simple requirement. Perhaps a business has decided to leverage Azure as their public cloud of choice but has a requirement to provide local access to a large file share within their facilities. Perhaps their WAN bandwidth to the public cloud is constrained or expensive, and they want to minimize any non-critical use of their WAN services during peak periods. Using Azure Stack Edge, they could expose a file share locally on the network that's backed to a storage account in the Azure public cloud, but only synchronize this data on a specified interval configured within the Azure portal.

Businesses consumption of Azure Stack Edge

Azure Stack Edge is consumed as a hardware-as-a-service solution. Businesses that are familiar with the consumption-based pricing of the Azure public cloud are already familiar with Azure Stack Edge's consumption model. Azure Stack Edge devices can be procured directly through the Azure Portal and businesses can pay for it as an operational expense billed monthly to their Azure subscription. When ordering an Azure Stack Edge device, shipping details are required, with most devices arriving within 2-3 weeks of order.

Final thoughts

Microsoft's Azure public cloud technology allows businesses to unlock the value of IT more quickly than could ever be done before, but there are also situations where WAN constraints, latency, or a general lack of connectivity to the public cloud require changes to how public cloud infrastructure is deployed.  Microsoft Azure and its Intelligent Edge offerings provide businesses with options about how best to leverage Microsoft Azure's capabilities in edge scenarios.

If you would like to discuss potential use cases for edge computing, we at World Wide Technology would love to help! Reach out to us to discuss your use case.