Edge Computing and the Towers of the Future

By Don Bishop, executive editor, assistant publisher, AGL Magazine

April 30, 2020

Power, space and transport are important factors that influence the strategy used to select macro tower sites and small cells as locations for edge computing infrastructure.

ExteNet Systems provides converged communication infrastructure-as-a-service. The company also designs, builds, owns and operates distributed networks both outdoors and indoors. The company’s customers are mobile network operators, service providers, enterprises, municipalities and venue owners. Outdoor and indoor distributed antennas systems (DASs), Wi-Fi access points and small cells are examples of what the company deploys.

Edge computing overcomes excessive latency that results when networks transfer data or queries far from the user, all the way to the cloud — a data center full of servers connected with the internet. “The natural solution is to bring computing, processing and storage closer to the users,” said Keyur Brahmbhatt, senior product manager at ExteNet. Speaking in January at the Seattle AGL Local Summit, Brahmbhatt said the

“Use cases, deployment models and business factors together are going to be important in making site selections.”

— Keyur Brahmbhatt, senior product manager at ExteNet Systems
Photo by Don Bishop

company uses two classifications of the edge cloud that brings cloud resources for analytics or computing processing close to the edge of the users.

Two Classifications

“One classification is network-based edge computing in which mobile network operators deploy edge computing as part of their network infrastructure,” Brahmbhatt said. “The second classification is premises-based edge computing in which an enterprise or venue owners deploy edge computing as if it were part of their IT infrastructure.”

Network-based edge computing encompasses cell towers and small cell hubs at utility or municipal poles, Brahmbhatt said. He said these become micro data centers, and because they are so close (just a hop away) from their mobile and nomadic users, they provide network operators with excellent locations for ultra-low-latency services. Examples he gave include immersive augmented and virtual reality (AR/VR), vehicle-to-everything (V2X) autonomous vehicles and edge analytics.

Mini Data Centers

At a greater distance are mini data center and local aggregation hubs for metro deployments, where data from access-based networks comes together (connected via fiber) for routing and switching, Brahmbhatt said. The mini data centers support real-time low-latency applications that have a little more latency tolerance, he said. As examples, Brahmbhatt mentioned the control and monitoring of automated guided vehicles at a seaport and super robots used in a distribution warehouse. He said edge data centers or edge computing procedures can manage those systems.

Machine learning, data analytics, virtual PCs (virtualization programs for Microsoft Windows) and 5G end-to-end network slicing are found at an even greater distance at regional data centers because they are not so sensitive to latency. Even further away, operators can use regional data centers to provide edge data centers for the support of latency-tolerant and non-real-time applications such as end-to-end (E2E) network orchestration, cloud-based services and content delivery network (CDN) and virtualized Evolved Packet Core (vEPC) solutions. “This is where ExteNet can also play a big role,” Brahmbhatt said.

You can expect to see enterprises and venue owners evolve existing IT infrastructure to support the low-latency services where 5G end-to-end network slicing are found and where virtual machine (VM)-based containers or architectures on generic hardware have been deployed, Brahmbhatt said. He said that ExteNet has been active in that business for several years, using its virtualized packet core to enable its Tier 3 mobility along with fixed wireless customers using the first ExteNet computing platform. Moreover, Brahmbhatt said, ExteNet has been active with enterprise and venue owners to migrate existing 4G network service to the 5G network.

Use Cases and Site Selection

When it comes to use cases and site selection, Brahmbhatt said that Extent starts with a survey.

“ExteNet actively participates with various industry forums active in small cell deployment. Small Cell Forum is a leading organization that works with the global operators to understand what the requirements are for small cell densification, and edge computing is a big part of it,” Brahmbhatt said. He said in 2017, the organization conducted a large survey to understand the key use cases and motivations for using edge computing.

Part of the information collected pointed to the importance of the ability to deploy small cells where opportunities can derive the most value, and that means enterprise customers, he said. For ExteNet, Brahmbhatt said, relevant enterprise sectors include commercial real estate, sports and entertainment, municipalities, health care organizations, hospitality and transit.

“Premises-based edge computing is important because of the need to collocate edge computing on a converged infrastructure where the traditional Internet Protocol (IP) local-area network (LAN) infrastructure exists,” Brahmbhatt said. “What that allows is a myriad of new use cases and deployment models.”

Previously, Brahmbhatt said, the radio access network (RAN) with its eNodeBs and gNodeBs was more distributed. With 5G, he explained, the RAN is split at the protocol layer. “You have your central unit and your distributed unit,” he said. “That central unit is now coming closer to the edge. This is your new centralized location. So you have the opportunity to virtualize the central unit and locate it at the edge of your premises. That is another opportunity to offer network-based service on edge infrastructure. It is network-as-a-service.”

As Brahmbhatt explained, infrastructure owners can offer centralized units. They can put them as virtual machines (VMs) on a virtualized VM-based platform or on a container-based platform. Linux Foundation’s edge project, Akriano, is developing the framework for edge computing use cases, he said.

At the other end is the core network, which Brahmbhatt said is typically more of a centralized platform, but he said that is changing.

“The 3rd Generation Partnership Project (3GPP) is splitting the core network at the control plane and at the user plane,” he said. “The user plane is coming closer to the edge. The user plane also can be virtualized and it can be at the edge. So you have the opportunity to provide the central unit for the RAN and your user plane for the core network.

“You can do local breakout and other network-based services, so that’s a new opportunity for edge-based network services,” Brahmbhatt added. “This is your network-as-a-service on top of your platform infrastructure-as-a-service. And then you have the opportunity to do platform-as-a-service. Essentially you have your edge domain name services (DNSs) (for resolving edge server names), edge policy enforcements, application discovery services and edge storage services. Value-adding services also include radio network information service for analytics of the underlying radio network, presence and location services (for detecting the presence of and possibly tracking the location of the user equipment (UE) devices), video compression services, video analytics services (such as facial recognition) and encryption services.”

Furthermore, Brahmbhatt said, with the service model called platform-as-a-service, the operator can add another layer with the edge-based applications that are specific to or tailored for the uses or devices at the edge. Examples he gave included augmented reality or virtual reality in a museum and robotic control in an industrial factory.

“All those services can yet be another layer of service on top of the edge computing infrastructure, so essentially new models and business cases will evolve,” Brahmbhatt said. “This is essentially the 3GPP 5G vision. As you go and explore these other opportunities beyond your typical tower site and the data center models, these are some of the new areas of opportunity. You just need to break the barriers.”

Brahmbhatt pointed out that in the classical cell tower case, beyond site selection, location is paramount because it must be within one hop to the mobile and nomadic user. He said ground space also is important for positioning servers, hardware and other infrastructure. It can be an existing space, a purpose-built space or an enclosure for edge-computing infrastructure.

Citing the need for adequate back-up, Brahmbhatt said edge computing equipment should have backup generators. He reminded the audience that the power requirements of 5G small cells are high. With the bandwidth requirements of 400 megahertz to 800 megahertz for 5G small cells and with multiple small cells and three-sector 5G antenna configurations, he said the power requirement can be as high as 2 kilowatts to 4 kilowatts per site.

With ExteNet Systems in the business of providing neutral-host service, Brahmbhatt said the company likes to collocate tenants. Having multiple operators, multiple technologies and multiple collocated small cells could extend power requirements to 10 kilowatts to 150 kilowatts per micro data center, he said.

Also required is adequate transport for backhaul and front-haul, Brahmbhatt said, explaining that very low-latency, high-bandwidth applications require a lot of fiber. He said that each cell site for a 5G location may require 12 to 24 strands of fiber and transport, and SFP, SFP+ and QSFP-type high-bandwidth-capacity transceivers. Power, space and transport are important factors that influence the site selection strategy, Brahmbhatt said.

“Use cases, deployment models and business opportunities together are going to be important in doing site selections,” Brahmbhatt said. “You need to make sure that the fundamentals and required essentials of site selections for next-generation sites are taken care of.”