Walking on the Edge
Between every two technologists there are three or more definitions of edge computing. There’s nowhere on the tech landscape today that creates more divergence of perspective as the term edge, and in my opinion, lack of crisp taxonomy limits our collective progress. I am, after all, a words person, and therefore am biased on the importance of coalescence on terms. We’ve seen this challenge before. Cloud comes to mind as a term that required ages to get crisp, and today some would argue that we’re still not unified on that definition. We are turning our focus at the TechArena to the edge with a span of compute, industrial, network and mobile edge and will be talking to industry icons and disruptors about how they’re defining their edge implementations and where market adoption is today for edge services. But before we get to all of that, I wanted to ground on my current definition of the edge in terms of what the industry is delivering today that is disrupting the technology landscape.
Let’s start by laying out some TechArena guardrails. The edge is not everything that is not in a data center. It is not all devices that are not directly human controlled (ie PCs, phones and smart watches). The edge is not a single place and it’s not a single thing…but the various edges do have some common characteristics that unify these edgy things as a unified commonality. And at this point of the diatribe, I’m going to promote the edge to the Edge…because there was use of the term edge in many corners of our industry before we started talking about the Edge. Those things may be part of the Edge…but they may not, and there is not a grandfathering clause that if you were once referred to as a part of the “insert term” edge you must be Edge.
With those guardrails, I’d like to explore why the focus on the Edge exists, in other words what problem the Edge is trying to solve for us. In listening to people discuss what they’re doing with Edge I think the three forces that created opportunity for Edge is as follows:
- Boatloads of data are growing on a distributed computing landscape.
- There are latency and data movement limitations in a traditional data center computing model where large compute capacity resides in a limited number of locations and network limitations bottleneck data movement.
- The operational efficiency and scale of cloud architectural models derived in data centers offer a way to unify computing across a growing diversity of compute environments.
If we combine these forces, together we can derive a definition of Edge:
A computing environment residing apart from a traditional data center location and following the core definitions of a cloud operational model to deliver efficient and fast digital services and/or data analysis.
Let’s test this definition against use cases from across the technology landscape.
Industrial Control
A manufacturing floor has deployed sensors and cameras to collect real-time production data to measure factory output and ensure factory safety compliance. This data collection is analyzed by Edge servers to ensure real-time control of the factory while sending summarized analysis to the organization’s data center. In this usage the servers as well as potential on-camera analytics are examples of Edge computing. Smart sensors that provision work across a sensor network may also be part of the Edge implementation within our taxonomy, and it’s the operational model of service provisioning, not the hardware’s existence, that would determine inclusion. Broader “collect only” sensors or “record only” cameras are part of the larger IoT network providing data collection for the Edge. Industrial Edge implementations will often tap 5G private networks for a mobile Edge computing solution within the factory environment – watch this space for more information to come about 5G private networks and mobile Edge.
An example of industrial Edge implementation is Worchester Bosch’s implementation of industrial Edge control of their UK boiler plant leveraging robotics, a sensor network of machine and collision sensors, and Edge server analytics controlled by a private 5G mobile network. This implementation has driven up factory output by 2% while increasing safety within the factory environment. Ericsson provides more information about this implementation on their site.
CDN (Content Delivery Networks)
Content delivery networks saw massive growth during the pandemic while we were all bound to our homes and are one of the fastest growing applications of Edge computing. In this case, an OTT provider deploys Edge servers close to customer locations to serve content to minimize latency and media consumption of core networks. These CDNs also provide Edge network security and microservices for collection of billing data. The CDN is functioning as part of the Edge per our taxonomy as it’s delivering cloud services outside of a traditional data center location and may be providing some analytics on customer traffic patterns to re-deploy most desired content to maximize customer viewing experience.
An example of a CDN in action is Netflix OpenConnect Edge network distributed to the >200 million customers streaming their content daily. This Edge implementation is matched with an AWS cloud backend where Netflix runs its data storage, customer analytics, recommendation engines, transcoding and more services which are not limited by latency requirements. The unified Edge to cloud solution ensures customers do not experience disruptions of service and Netflix is able to continue analyzing customer consumption data to deliver meaningful content recommendations and move content across their network to serve optimal content to customers.
VRAN (Virtualized Radio Access Networks)
In some ways, VRAN is the holy grail of Edge implementations simplifying and speeding radio access network services (those things that keep you and I reliably connected to mobile networks) while unifying core, Edge and RAN in a common platform. To understand VRAN we need to introduce some additional terms including centralized unit (CU) an access control point which provides RAN oversight and non-real time data processing, distributed unit (DU), controlling lower-level protocols including MAC layer control through real time processing, and remote radio unit (RRU), providing physical layer transmission and reception. By virtualizing these three functions, or containerizing them in a cloud-native implementation, telecom providers are able to run their radio access networks on standard server hardware and reduce or eliminate costly proprietary solutions.
While there are various configurations of VRAN solutions, some featuring open interfaces for the RRU, others maintaining priority hardware for this portion of the RAN, the cost savings of these virtualized solutions are driving mass disruption as providers move to 5G networks and seek cloud native service to deliver the full promise of the 5G standard. To do this they require the cloud operational control of services from core to Edge including the RAN. The expected investment in this space is staggering with over $550 billion in collective VRAN investment in the next few years alone. Watch this space for more information on the state of VRAN and to hear about progress in VRAN solution deployments in the weeks ahead.
These three examples are just scratching the surface on Edge implementations, and the TechArena is exploring other use cases to bring into the discussion. We’re also looking to hone this definition with the industry in the months ahead as we begin discussions with innovation experts from across the multi-faceted Edge landscape. As always, thanks for engaging - Allyson
Between every two technologists there are three or more definitions of edge computing. There’s nowhere on the tech landscape today that creates more divergence of perspective as the term edge, and in my opinion, lack of crisp taxonomy limits our collective progress. I am, after all, a words person, and therefore am biased on the importance of coalescence on terms. We’ve seen this challenge before. Cloud comes to mind as a term that required ages to get crisp, and today some would argue that we’re still not unified on that definition. We are turning our focus at the TechArena to the edge with a span of compute, industrial, network and mobile edge and will be talking to industry icons and disruptors about how they’re defining their edge implementations and where market adoption is today for edge services. But before we get to all of that, I wanted to ground on my current definition of the edge in terms of what the industry is delivering today that is disrupting the technology landscape.
Let’s start by laying out some TechArena guardrails. The edge is not everything that is not in a data center. It is not all devices that are not directly human controlled (ie PCs, phones and smart watches). The edge is not a single place and it’s not a single thing…but the various edges do have some common characteristics that unify these edgy things as a unified commonality. And at this point of the diatribe, I’m going to promote the edge to the Edge…because there was use of the term edge in many corners of our industry before we started talking about the Edge. Those things may be part of the Edge…but they may not, and there is not a grandfathering clause that if you were once referred to as a part of the “insert term” edge you must be Edge.
With those guardrails, I’d like to explore why the focus on the Edge exists, in other words what problem the Edge is trying to solve for us. In listening to people discuss what they’re doing with Edge I think the three forces that created opportunity for Edge is as follows:
- Boatloads of data are growing on a distributed computing landscape.
- There are latency and data movement limitations in a traditional data center computing model where large compute capacity resides in a limited number of locations and network limitations bottleneck data movement.
- The operational efficiency and scale of cloud architectural models derived in data centers offer a way to unify computing across a growing diversity of compute environments.
If we combine these forces, together we can derive a definition of Edge:
A computing environment residing apart from a traditional data center location and following the core definitions of a cloud operational model to deliver efficient and fast digital services and/or data analysis.
Let’s test this definition against use cases from across the technology landscape.
Industrial Control
A manufacturing floor has deployed sensors and cameras to collect real-time production data to measure factory output and ensure factory safety compliance. This data collection is analyzed by Edge servers to ensure real-time control of the factory while sending summarized analysis to the organization’s data center. In this usage the servers as well as potential on-camera analytics are examples of Edge computing. Smart sensors that provision work across a sensor network may also be part of the Edge implementation within our taxonomy, and it’s the operational model of service provisioning, not the hardware’s existence, that would determine inclusion. Broader “collect only” sensors or “record only” cameras are part of the larger IoT network providing data collection for the Edge. Industrial Edge implementations will often tap 5G private networks for a mobile Edge computing solution within the factory environment – watch this space for more information to come about 5G private networks and mobile Edge.
An example of industrial Edge implementation is Worchester Bosch’s implementation of industrial Edge control of their UK boiler plant leveraging robotics, a sensor network of machine and collision sensors, and Edge server analytics controlled by a private 5G mobile network. This implementation has driven up factory output by 2% while increasing safety within the factory environment. Ericsson provides more information about this implementation on their site.
CDN (Content Delivery Networks)
Content delivery networks saw massive growth during the pandemic while we were all bound to our homes and are one of the fastest growing applications of Edge computing. In this case, an OTT provider deploys Edge servers close to customer locations to serve content to minimize latency and media consumption of core networks. These CDNs also provide Edge network security and microservices for collection of billing data. The CDN is functioning as part of the Edge per our taxonomy as it’s delivering cloud services outside of a traditional data center location and may be providing some analytics on customer traffic patterns to re-deploy most desired content to maximize customer viewing experience.
An example of a CDN in action is Netflix OpenConnect Edge network distributed to the >200 million customers streaming their content daily. This Edge implementation is matched with an AWS cloud backend where Netflix runs its data storage, customer analytics, recommendation engines, transcoding and more services which are not limited by latency requirements. The unified Edge to cloud solution ensures customers do not experience disruptions of service and Netflix is able to continue analyzing customer consumption data to deliver meaningful content recommendations and move content across their network to serve optimal content to customers.
VRAN (Virtualized Radio Access Networks)
In some ways, VRAN is the holy grail of Edge implementations simplifying and speeding radio access network services (those things that keep you and I reliably connected to mobile networks) while unifying core, Edge and RAN in a common platform. To understand VRAN we need to introduce some additional terms including centralized unit (CU) an access control point which provides RAN oversight and non-real time data processing, distributed unit (DU), controlling lower-level protocols including MAC layer control through real time processing, and remote radio unit (RRU), providing physical layer transmission and reception. By virtualizing these three functions, or containerizing them in a cloud-native implementation, telecom providers are able to run their radio access networks on standard server hardware and reduce or eliminate costly proprietary solutions.
While there are various configurations of VRAN solutions, some featuring open interfaces for the RRU, others maintaining priority hardware for this portion of the RAN, the cost savings of these virtualized solutions are driving mass disruption as providers move to 5G networks and seek cloud native service to deliver the full promise of the 5G standard. To do this they require the cloud operational control of services from core to Edge including the RAN. The expected investment in this space is staggering with over $550 billion in collective VRAN investment in the next few years alone. Watch this space for more information on the state of VRAN and to hear about progress in VRAN solution deployments in the weeks ahead.
These three examples are just scratching the surface on Edge implementations, and the TechArena is exploring other use cases to bring into the discussion. We’re also looking to hone this definition with the industry in the months ahead as we begin discussions with innovation experts from across the multi-faceted Edge landscape. As always, thanks for engaging - Allyson
