Edge computing is one of the latest and hottest buzzwords in the wireless industry. Already, AT&T is testing edge computing designs and Verizon has gone as far as to coin the term “intelligent edge network” for its latest upgrade efforts.
The noise behind edge computing is not necessarily a surprise; the technology is being pushed by three interrelated factors:
1. Data processing needs to happen closer to the end user to lower latency speeds.
2. Network operators must densify their networks as 5G approaches.
3. The virtualization of network elements reduces costs and encourages shared infrastructure.
After all, the argument goes, applications such as autonomous vehicles, virtual reality, robotic surgery and myriad others simply are not possible using older, slower network architectures.
"To get to the 5G performance goals of low latency, … you need to push content and apps as close to the radio as you can," said longtime industry analyst Iain Gillott, president of consulting firm iGR. "The idea is that to send data all the way back to radio and backhaul packet to a cloud processor somewhere takes too long. It's as simple as that."
But perhaps the most interesting element of the industry’s push to edge computing is that it’s creating new opportunities for both new and established market competitors: "You are going to see new players," said John Baker, Mavenir Systems’ SVP of business development. (Mavenir is one of those companies hoping to cash in on the move to edge computing.)
Continued Baker: "You are going to see customer hardware manufacturers being replaced by OEM computer manufacturers such as Dell, HP and Intel. It's the cannibalization of the telecom, which used to be dominated by customized telecom hardware. This hardware will be cannibalized by off-the-shelf computing hardware and off-the-shelf computing functions. There will be opportunities for a new set of players that are nimble and have no legacy to project."
New players and approaches
To be clear, computing at the edge is not a tweak; instead, it’s one element of a mobile network transformation. Ed Chan, Verizon’s SVP of technology strategy and planning, described the technology’s complex issues and deep possibilities late last year at an investor conference. Specifically, he said changes include network virtualization, creation of software "control point[s]" to manage that network, the sharing of transport across multiple networks, and the creation of cloudlike environments at the edge and the core.
In short, Verizon's version of edge computing is centrally managed virtual infrastructure supporting core and local clouds. This obviously will call for lots of storage and computing in new and in some cases cramped, nontraditional quarters.
Said Chan, according to a transcript: "[W]hat gets really exciting is when you get to that 5G place and you add this kind of edge-cloud capabilities, you started seeing the fact that, wow, would it be interesting that it feels like the cloud is right in your back pocket when the latencies got short, right, when the bandwidth is so high."
However, the blueprint for how the wireless network of the future will look is far from set. “There is no single best strategy for compute at the edge,” wrote Amazon Web Services IoT VP Dirk Didascalou in an emailed response to FierceWireless. “Rather, manufacturers and enterprises make use-case specific decisions to balance the low-latency of local execution with the elastic resources of the cloud.”
And it’s that uncertainty that may be driving the interest among vendors in edge computing. Indeed, the landscape is full of interesting ideas from smart startups and savvy legacy companies that understand that networks are changing at a very basic level. Here’s a look at some—but certainly not all—of the players on the scene:
Vapor IO: This startup came out of stealth in March 2015. CMO Matthew Trifiro said it is creating a suite of colocation edge data centers and is partnering with tower giant Crown Castle on their joint venture Project Volutus to embed them in carriers’ RAN. The company’s long-term goal is to create citywide edge computing environments it calls the Kinetic Edge. Basically, the company’s concept is a no-touch colocation facility for compute, storage and networking equipment.
Vapor IO said it is seeing demand from telcos that want to virtualize their networks, content delivery networks that want to store content near end users, web-scale companies that want to own their own infrastructure, and public cloud providers. The company’s systems are operational now in Chicago and Austin, with wider deployments coming later this year.
Vapor IO ultimately is targeting a wide range of customers including land and building owners, colocation providers, public cloud providers, content delivery networks and enterprises.
EdgeMicro: Startup EdgeMicro has not yet celebrated its first birthday. It offers what in essence are mobile colocation-in-a-box solutions. The company’s containerized mobile data centers, built in association with Schneider Electric, feature six racks that are rated at 48 watts. And data traffic can move between a cell tower and EdgeMicro's modular data center via an IP proprietary protocol called Tower Traffic Xchange. These data centers, said CTO and co-founder Anton Kapela, can cut huge amounts of time and work out of traditional backhauling.
EdgeMicro data centers can be deployed at towers, metro fiber splice junctions, rooftops, parking lots and elsewhere. The company said that not relying on a single place to deploy its equipment encourages options with a variety of players, from mobile network operators to CDNs, content owners and others.
EdgeMicro has announced an ambitious rollout plan, though the details are fuzzy. The company plans to deploy sites in cities including Portland, Sacramento, El Paso and Memphis. The company’s target customers include service providers, mobile network operators and tower and facility owners.
Rigado: Startup Rigado, which has been around since 2010, offers a variety of edge solutions that combine computing, multiprotocol connectivity and security. The company’s goal is to create smart buildings, retail locales and hospitality sites.
Rigado’s strategy is to provide flexibility. The company said it can support use cases ranging from real-time smart city functions such as traffic control to managing a universe of IoT devices that need battery life measured in years. The way in which each is handled is significantly different, said Chief Revenue Officer Kevin Tate. The company's Cascade Edge-as-a-Service costs $9 per active gateway per month.
Rigado offers connectivity as a service in a variety of areas, including asset tracking, smart lighting and others. Available on a subscription basis, it provides connectivity across networks like Wi-Fi and cellular.
The edge computing element of the company has about 350 end-user company customers. “We zoomed in on a particular subset, which is the edge infrastructure for commercial IoT use cases in which we’re trying to create repeatability in the connectivity, computing and cloud integration,” Tate said.
Saguna: Saguna, which launched 10 years ago, offers hybrid private/public cloud services to mobile operators, according to Marketing Director Tally Netzer. The company's software operates at tower aggregation points.
Saguna’s software identifies data streams from client companies, which it processes in a private cloud. In addition to avoiding latency issues, the company said that approach enables its client customers to enjoy the benefits of combining a private cloud and public network. Saguna is testing its offering with carriers in the United States and elsewhere.
AWS Greengrass: AWS Greengrass, from web giant Amazon, offers edge computing services to end-user devices on a subscription basis. The services run compute, messaging, data caching and sync capabilities at the edge, AWS’ Didascalou noted. The platform can be used for writing and testing code, routing local messages without internet access, and syncing with the AWS cloud when connectivity is restored. The platform also offers a machine learning interface.
Didascalou said the company’s capabilities are illustrated by networked traffic lights: These cameras can collect as much as 10 GBs of data a day, but as little as 1% of that data is useful. Thus, sending all that data to a distant data center—where most of it will be discarded—is wasteful.
"With AWS Greengrass, they have the capability to process and perform [machine learning] inference on the data locally," Didascalou explained. "For example, customers using AWS Greengrass ML Inference can identify a specific object in an image and then only upload this information to the cloud, instead of the entire video stream. This saves the customer money and time."
Nokia: Jane Rygaard, head of marketing for Nokia's mobile networks business, said the company has been working on a new version of mobile edge computing since Nokia’s introduction of Liquid Computing in 2012.
The company’s concept is to create a distributed architecture in which capacity is fluidly distributed among centralized and edge computing elements. Workloads would be distributed according to the needs of the work being done, and an orchestration layer would coordinate computing between the core and edge computing elements.
Nokia’s basic concept is to right-size the cloud so that it can provide the needed functionality without overprovisioning. The company’s Open Edge Cloud Solution has been deployed on its existing AirFrame architecture. The next step—which Rygaard said is expected this quarter—is the AirFrame Open Edge Server. The difference between the two, Rygaard said, is that the new server will have the ability to scale down. The ability to not overprovision is important because powering edge data centers is trickier than consolidated facilities.
“The edge clouds have to be power efficient,” Rygaard explained.
Mavenir Systems: Mavenir was formed in February 2017 by the combination of Mitel Mobility and Ranzure Networks. The company helps network providers by enabling them to use the cloud and operate with "webscale" speed and agility.
Mavenir’s Baker explained that as much as 80% of the data on mobile networks is video, and that trafficking that data through the mobile network core is expensive. Baker said that offloading this data from the packet core to the cloud enables operators to defer investment in their core for two years.
Mavenir said it can do this through its virtualized Media Breakout Controller (vMBC), which it introduced in January. The company’s vMBC is a white-box appliance that service providers can use to host virtual routing, firewall and user plan data offload, the company said. In other words, Mavenir can move data flows to the cloud where it is feasible.
Edge basics
There clearly are both "sexy" and "boring" ways in which these new capabilities can be harnessed. On the exciting side, high-speed 5G services powered by edge computing designs can support autonomous vehicles, gaming, augmented and virtual reality, and other high-profile applications.
But, as Gillott noted, edge computing can also support "boring" applications, like video storage. Consider a retailer with 1000 locations nationwide that wants to show training videos to each new hire, periodically upgrade and manage applications, and perform similar tasks that currently require communications with the corporate data center. "What can be done is put in edge computing into the stores," Gillott said. "All the content is locally cached overnight. It's all sitting there when they need it."
Apart from the applications, edge computing nonetheless appears to be a trend set to overtake mobile network operators and their suppliers. The resulting network speeds and flexibility clearly can power new services that could lead to more revenues. But, getting to that point is unclear at best considering the variety of approaches to edge computing advocated by providers both big and small, and both new and established.