It might surprise you to learn that the big three public clouds – AWS, Google Could Platform, and Microsoft Azure – are all starting to provide edge computing capabilities. It’s puzzling, because the phrase “edge computing” implies a mini datacenter, typically connected to IoT devices and deployed at an enterprise network’s edge rather than in the cloud.
The big three clouds have only partial control over such key edge attributes as location, network, and infrastructure. Can they truly provide edge computing capabilities?
The answer is yes, although the public cloud providers are developing their edge computing services via strategic partnerships and with some early-stage limitations.
Cloud-based edge computing offerings are a clear sign that the boundaries between public cloud, private cloud, and edge computing are blurring. The unifying goal is to provide businesses and architects with a range of choices based on the type of workload and its performance, reliability, regulatory, and safety requirements.
Unfortunately, a glut of new options always means new jargon and branding, so we’ll need to do a fair bit of demystifying as we sort through the big three cloud offerings for edge computing. Before jumping in, though, let’s start with a quick primer on some key edge computing architecture considerations.
Understanding edge computing requirements and architectures
First and foremost, engineering teams must understand the edge computing requirements. Connecting a globally dispersed network of inexpensive sensors that generate a few terabytes of data daily has different computing requirements than servicing a dozen factory floors with an array of video sensors processing petabytes of data in real-time. The architecture must address the specific data processing, analytics, and workflows needed.
Then, just as important, consider the regulatory, security, and safety requirements. Medical devices deployed at hospitals or controllers for autonomous vehicles both capture and process highly personal, life-critical information. Reliability and performance demands should dictate the location, network, security, and infrastructure requirements.
Understanding these requirements helps architects determine where to locate edge computing infrastructure physically, what types of infrastructure are required, the minimal connectivity requirements, and other design considerations.
But the unique benefit public cloud edge computing offers is the ability to extend underlying cloud architecture and services, particularly for customers already heavily invested in one public cloud or another. Do architects and developers want to leverage AWS, Azure, or Google Cloud services deployed to the edge? That’s what the public clouds are betting on – and they are also considering 5G-enabled mobile applications that require low-latency data and machine learning processing at telco endpoints.
With these questions in mind, here’s an overview of what the three major public clouds provide.
Extend to Azure Edge Zones with Azure Stack
Azure is betting that architects and developers want to focus on the application and less on the infrastructure. Azure has three options that enable a hybrid edge, where architects can leverage 5G networks and deploy data processing, machine learning models, streaming applications, and other real-time data-intensive applications optimally.
- Azure Edge Zones are managed deployments of the Azure stack that can be purchased through Microsoft and are currently available in New York, Los Angeles, and Miami.
- Microsoft partnered with AT&T to offer Azure End Zones with Carrier in a number of locations, including Atlanta, Dallas, and Los Angeles. This option is best for 5G-enabled mobile applications that require low-latency data processing or machine learning capabilities.
- Lastly, businesses can also deploy a private Azure Edge Zone. Microsoft has partnered with several data center providers that enable this capability.
These options provide location choices and network flexibility, while Azure Stack Edge brings Azure computing and services to the edge. Azure Stack Edge is a 1U, 2×10 Core Intel Xeon, 128GB appliance that can be configured with containers or VMs and managed as a Kubernetes cluster of appliances. This model is optimized for machine learning and IoT applications.
Microsoft also offers Azure Stack HCI, a hyperconverged infrastructure for modernizing data centers, and Azure Stack Hub for deploying cloud-native applications.
Like other cloud services, Microsoft sells Azure Stack Edge by subscription, with costs calculated by the utility. Microsoft manages the device and offers 99.9 percent service level availability.
Extending AWS services from 5G devices to large scale analytics
AWS has a similar set of offerings to distribute AWS services to edge data centers and telco networks.
- AWS is starting to support edge data centers with AWS Local Zones that are currently available only in Los Angeles.
- AWS Wavelength is designed for low-latency applications running on 5G devices, including connected vehicles, AR/VR applications, smart factories, and real-time gaming.
- AWS is partnering with Verizon to deliver AWS Wavelength, which is currently available in Boston and the San Francisco Bay area.
AWS offers two flavors of edge infrastructure, starting with the AWS Snow line of appliances. AWS Snowcone is the smallest appliance with two vCPUs and 4GB that is primarily used for edge data storage and transfer. Memory-intensive data processing and machine learning applications deployed to the edge likely require AWS Snowball Edge that comes in storage and compute-optimized models with up to 52 vCPUs and 208GB. For the largest scale applications, AWS Outposts are 42U racks deployed to data centers for running different EC2 instance types, containers (Amazon ECS), Kubernetes (Amazon EKS), databases (Amazon RDS), data analytics (Amazon EMR), and other AWS services.
Google lags as all three clouds vie for the edge
Just as Google occupies third place in the public cloud wars, it’s also trying to catch up with its edge offerings. Google’s most recent announcements include Anthos at the Edge, collaborations with AT&T on 5G connectivity, and the Google Mobile Edge Cloud. The offering is part of Anthos, a hybrid and multi-cloud application modernization platform enabling businesses to deploy applications on GCP and in the data center.
The public cloud vendors all recognize that the next wave of innovation is coming from the intersection of IoT, 5G, and machine learning analytics deployed at the edge. They’re not going to let infrastructure and data center companies like Dell or HPE dominate this new market without a fight, so their answer is to bring their cloud platforms, containers, orchestration, and services to edge data centers and telco endpoints. And they’re not doing this alone: The public clouds are partnering with telcos, infrastructure providers, and major service providers to enable their offerings.
But these are early days for public cloud edge computing solutions. The fact that the big three clouds are getting serious about edge computing only serves to underscore the promise of the edge frontier. Whether enterprises choose public cloud edge solutions or opt to build their own infrastructure, network, and computing platforms, few will want to be left out of the growing wave of edge innovation.