Demystifying the unspoken plurality of edge

Next-gen wireless technologies, such as 5G and Wi-Fi 6E, promise to unlock ubiquitous high-speed connectivity, which means conversations about edge technology continue to gain steam among industry pundits. The majority of these conversations fail to acknowledge the plurality — and ambiguity — that comes with the terminology. After all, the term edge means something completely different to a wireless network engineer versus a data center admin.

The device edge and telco network compute edge are both considered edge computing but warrant two entirely different conversations around use cases, applications and workloads. Because the term is so broad, conversations about the edge quickly enter muddy water and become counterproductive.

The three frequently discussed edge computing types are the following:

  1. Device edge. An end-user device that generates, processes and transmits data.
  2. Network edge. A point in network infrastructure where compute can be performed.
  3. Cloud edge. A server that provides distributed compute via a centralized offering.

As organizations continue to adopt distributed networks, the many types of edge only become more critical. Therefore, it’s imperative that technology experts add more distinction and specificity when talking about the many facets of this important technology.

Device edge

End-user devices are literally located at the end of the network, where the application is used. These devices collect data and transmit it to the centralized compute infrastructure. They also process, analyze and perform necessary actions on locally collected data. Since 2010, devices have become much smaller with significantly more local compute power, which enables devices to perform heavier and more sophisticated workloads.

Apple and Android already offer on-device machine learning kits, real-time augmented reality, facial recognition and live camera enhancement in a push to help move more intelligence closer to this edge.

The ability to move more computing power to edge devices is increasingly important as the amount of generated data continues to exponentially grow. By 2025, global data creation is projected to grow to more than 180 zettabytes, which is 10 times as much data as was produced just five years ago, a Statista report noted.

It is impractical to process and store such large volumes of data in a centralized cloud or data center, especially for time-sensitive processes that require intensive computations. Edge devices offer an effective and cost-efficient way to process and store data in a more distributed manner.

Consider the significance of the device edge network for autonomous vehicles. Despite what a 5G marketer would say, the vehicle’s on-device workloads are significantly more important than any network, telco or cloud edge workloads. The network can add tremendous value to the self-driving ecosystem, but the device edge workloads are truly critical.

Network edge

There are many embedded apps in network infrastructure that perform a variety of functions, such as data pre-processing and filtering or security service chaining. But even the term network edge means different things to different people. There are three common meanings: the telco network edge, the WAN edge and the on-ramp edge.

Because a great deal of the edge conversation relates to 5G, the telco network edge is what might first come to mind. The telco edge is composed of cell towers and equipment closets that house the telecom service provider’s distributed systems. Because of the telco network edge’s geographical distribution, it represents a footprint of facilities where compute is hosted near the mobile user to run workloads for latency-sensitive applications.

The WAN edge is another essential network edge layer because it represents both the connected perimeter of a site, and it provides the path to access cloud-hosted applications. The WAN edge is a crucial security boundary, but it also plays a key role in application optimization and delivery for all off-premises workloads.

Finally, the device on-ramp edge consists of Wi-Fi access points, Ethernet switches or private 5G small cells; it represents the first leg of connectivity. The on-ramp edge is pivotal both in terms of access control and policy, but it also provides a distributed compute platform for lightweight edge services, like security or IoT applications. This edge type also bears primary responsibility — and blame — for device connectivity and service continuity.

Cloud edge

A common misconception that continues to crop up in edge literature is that cloud and edge represent two fundamentally different technologies, which may or may not be true.

Most times, when people say cloud, they refer to a centralized compute platform in a remote, third-party data center. Cloud just refers to a specific way to pool compute and storage resources and make them available via a pay-per-use model.

When people compare cloud and edge technologies, they are often just comparing centralized and distributed computing models, both of which serve different purposes. That’s why something like cloud edge can sound like an oxymoron but is a useful tool for specific use cases.

Cloud edge refers to a distributed server or server clusters that provide distributed compute and centralized oversight and/or orchestration via cloud data center. The cloud edge essentially takes advantage of the software tools and architectures of the cloud but moves the actual software or application closer to the device edge.

This technology gives enterprises an option to have a small cloud instance in their private data center. If companies can locally store and process data, IT teams can deliver applications with lower latency and potentially with a tighter security boundary within the perimeter of the operator. This is important for a forthcoming generation of compute-intensive and latency-sensitive applications, such as streaming, online gaming and data pre-processing with anonymization.

About the author
Marcus Burton is a Wi-Fi and networking veteran who’s currently working at Extreme Networks as a cloud and wireless architect in the office of the CTO with a focus on solving customer headaches using data science, machine learning and AI. He has written books, white papers, blogs and exams as the technology and product lead at CWNP and also previously worked at Ruckus Wireless, where he held roles in product leadership.

Leave a Reply