Menu

EDGE FEATURE NEWS

Thick or Thin Edge: What's the Difference?

By Special Guest
Jim White, CTO, IOTech
June 12, 2020

One of the current trends in edge computing is to divide edge computing between “thin” and “thick”. Taking a page from client computing (thin vs thick clients), the idea is that edge computing can be broken down along lines of whether to distribute or centralize the computing in an edge deployment.

In thick edge computing, there is a move to a more distributed and decentralized computing approach. The sensors are often connected to compute nodes running edge applications close to where the data is collected, filtered, aggregated and batched, etc. From there, the data is then sent to cloud or enterprise data center environments. In thick edge computing, these local compute nodes (typically referred to as edge devices, edge gateways or edge computers) often provide localized intelligence to act quickly (in a low latency fashion) on the data coming in from physical sensing devices. Because of their processing and storage capability, thick edge computing nodes are often able to operate independently for periods of time.

In thin edge computing, less processing occurs at the point of sensor data collection. The sensed data is collected and passed back to more centralized processing where it is explored and acted upon. The centralized processing may happen in the cloud or it could occur in private data / processing centers (known as “on-prem” environments). There is minimal processing or intelligence happening at the thin edge. There may be little or no data filtering, aggregation, etc. happening in a thin edge deployment. In fact, depending on the sophistication and connectivity capability with the physical sensing equipment, additional computing platforms (e.g. gateways) may not even be needed in thin edge compute deployments.

In some environments, organizations may have to institute a simple edge node just to be able to connect legacy sensing equipment to the cloud or data center. However, in these environments, there is no application or intelligent processing in these nodes. In this sense, even though there is a more localized compute platform, it is behaving more like a thin edge device.

Thin / Thick Spectrum

While the industry often makes hard delineations, the line between what is a thick and thin edge can blur. Thin and thick labels treat the edge host platform and software application(s) – that is the processing and intelligence – as if the two were fused.

In reality, thin and thick edge is not discrete. There is more of a spectrum of solutions that tend to thin or thick based on resource availability, use case need, and software deployment and management concerns at the farthest reaches of the edge.

Forgive the pun, but there are many edge cases in thin / thick edge compute today that can defy labeling. For example, many modern security cameras (a physical sensing device) are becoming very sophisticated. In addition to the camera, many come equipped with other sensors (temperature, audio, light detection, etc.) and have onboard analytics applications. As an example, some cameras are able to run algorithms to automatically count the number of people seen. If networked directly to the cloud or on-premise data center, would you classify these as thin or thick edge physical sensing devices? Probably thick - given the intelligent processing at the edge, but when disconnected, they behave more like thin edge devices – unable to operate independently or provide any value outside of the backend systems to which they are connected.

Thin / Thick Edge Compute

As in thin vs thick client computing, there are pros and cons to thin vs thick edge computing. Low latency decision making, reduced data transportation and storage costs, ability to operate independently and disconnected for short periods of time, ability to address legacy equipment not easily connected directly to the cloud, and keeping the data at its origin (avoiding security and governance issues) are all benefits of thick edge computing. The ability to address scale, reduce maintenance issues in remote locations, faster/easier solution deployments (providing better agility), and ability to use more traditional IT enterprise solutions are some of the benefits of thin edge computing.

Keep in mind too, that there is a tipping point (albeit unclearly marked) where edge computing stops and normal distributed or enterprise computing begins.

Alternatives

You will find alternative names for thick and thin edge computing. For example, some analysts refer to “device” and “cloud” edge instead of thick and thin edge computing. Some vendors have referred to the thin edge or cloud edge as “sensor to cloud” computing. Individual industries such as telecommunications define their own “telecom edge” which refers to a place for these companies to transform mobile networks and accelerate the deployment of workloads from public clouds to the networks much closer to customer premises.

Conclusion

In summary, the distinction between “thick” and “thin” edge usually comes down to where the data processing and intelligence is located. However, this is not always the case and there is a spectrum between the two. In practice, where the boundary lies between the two definitions is more nuanced. As edge computing develops, industry will refine the definition of the industrial edge and new terminology will no doubt emerge and be added to the edge computing taxonomy.

About the author: Jim White is CTO at IOTech and has over 25 years of experience in software development for IoT Edge systems, enterprise application integration and mobile applications. Most recently, Jim was a Distinguished Engineer and Director of the IoT Platform Development Team within the IoT Solutions Division of Dell Technologies, where he was the chief architect for Dell's largest open source effort to date, EdgeX Foundry. EdgeX is an open framework for building industrial IoT Edge computing systems and is now a Linux Foundation (LF) Edge project. Jim currently serves as Vice Chair of the EdgeX Technical Steering Committee.




Edited by Ken Briodagh
SHARE THIS ARTICLE
Related Articles

Real World Use Cases for Intelligent Edge Infrastructure

By: Ken Briodagh    4/26/2022

The Intelligent Edge is being integrated into nearly every industry and now that aging infrastructure is being updated all over the world, Intelligent…

Read More

Are Energy and Utility Companies in Position to Generate New Revenues through 5G Services with Existing Physical Infrastructure?

By: Matthew Vulpis    4/11/2022

The recent, rapid rise of digitalization has brought with it a vast number of devices and applications revolutionizing the way daily processes get don…

Read More

Telco Global Alliance Introduced at Mobile World Congress: Linux Foundation Unveils CAMARA

By: Juhi Fadia    2/28/2022

In a surprise move that had everyone buzzing at Mobile World Congress in Barcelona today, the Linux Foundation, in collaboration with the GSMA, introd…

Read More

Supermicro Delivers Intelligent Edge

By: Maurice Nagle    2/17/2022

Supermicro announced the SuperEdge server, developed for environments demanding high-density compute and I/O in a small form factor. Optimized for Edg…

Read More

Industrial IoT Evolution at the Edge Picks Up the Pace with IOTech's Edge Xpert 2.1 Supporting EdgeXFoundry

By: Special Guest    2/3/2022

With the transformative influence of IoT innovation over various industries, there have been a plethora of new solutions pioneered that have evolved h…

Read More