Google Launches Mobile Edge Cloud with AT&T

Depending on how you wish to view it, Google’s new Global Mobile Edge Cloud strategy, which will deliver a portfolio and marketplace of 5G solutions built jointly with telecommunications companies, is about as good a role as connectivity providers likely can hope for, or stark recognition that a real estate play was always going to be the role service providers would assume. 

Google Cloud and AT&T announced they will work together to “help enterprises take advantage of Google Cloud’s technologies and capabilities using AT&T network connectivity at the edge.”

One might translate that as “computing at the edge will be supplied by Google, racks, energy, cooling and security at the edge by AT&T,” albeit with access using the low-latency 5G network. 

The two companies are testing a portfolio of 5G edge computing solutions for industries, such as retail, manufacturing, and transportation, that bring together AT&T’s network, Google Cloud’s leading technologies including AI/ML and Kubernetes, and edge computing to help enterprises address real business challenges, Google says. 

Google Cloud also announced Anthos for Telecom, which will bring its Anthos cloud platform to the network edge, allowing telecommunications companies to run their applications wherever it makes the most sense. 

Google also might partner with telecommunications companies to rapidly enable a global distributed edge by lighting up thousands of edge locations that are already deployed in telecom networks. The language is “can do so,” rather than “will do so.”

So far, this sort of collaboration, as well as Verizon's working with Amazon, tend to suggest that edge computing mostly will be supplied by the hyperscalers, while telcos mostly contribute edge computing facilities, while locking down some portion of the busines and enterprise 5G connections market.

That might realistically be about the best general outcome telcos can hope for. The danger is that they, once again, find they cannot move much beyond "connectivity" as their role in the ecosystem. There might be some incremental revenue created by supplying edge computing real estate.

But that will not move the revenue needle very much. It is hard to grow beyond dumb pipe. Edge computing now is shaping up to be the latest example of that reality.

Walmart Expects Edge Computing and 5G to Create New Lines of Business

Walmart believes 5G and edge computing might be enablers of a new healthcare capability inside its retail stores. 

The company hopes it can create healthcare delivery services by using 5G and edge computing to connect patients in-store with remote medical professionals, in addition to having in-store clinic personnel, for example. 

In addition, third party ability to use Walmart 5G, edge computing and logistics will create additional revenue streams from logistics as a service or computing as a service sales, much as 

Amazon Web Services now drives profit for Amazon.

At the same time, the edge computing and 5G should allow Walmart to support its internal operations with more-extensive video surveillance, perhaps video point of sale use cases and inventory monitoring. 

Walmart operates some 4,700 U.S. stores, many of which are “supercenters” of around 180,000 square feet offering 100,000 products. They are often open 24 hours and are essentially community gathering spots. Walmart estimates that 90 percent of the U.S. population lives within 10 miles of a Walmart location.

Is Walmart Looking at Some Forms of 5G Support for its Edge Operations?

Walmart has been in talks with some mobile operators about installing 5G antennas on Walmart store roofs. Precisely what the perceived advantage might be is not clear, but if Walmart also operates edge computing facilities open to third party customers, it might also want to ensure that  ultra low latency 5G is available to those edge computing customers.

Walmart might also be looking at millimeter wave connectivity for shoppers inside its store locations, perhaps to enable video-heavy applications inside the stores. 

Walmart also plans to build edge computing facilities in its stores, not only for its own use, but as a commercial service available to third party customers. Third parties also will be allowed to buy use of the warehouses and shipping services.

Global Carriers Work to Develop Common Edge Computing Framework

China Unicom, Deutsche Telekom, EE, KDDI, Orange, Singtel, SK Telecom, Telefonica and TIM have joined forces, with the support of the GSMA, to develop a multi-access edge computing platform that is interoperable. 

The platform, to be developed in 2020, will make local operator assets and capabilities, such as latency, compute and storage available to application developers and software vendors enabling them to fulfil the needs of enterprise clients.

There are other interoperability efforts also underway, including 3GPP efforts, in addition to work by ETSI. 

In the end, standards are not a business model, but an enabler of business models. Some note that the edge is a possible business battleground, as many in the ecosystem hope to capitalize on the opportunity as providers of computing as a service, computing platform as a service or colocation. 

Hyperscale cloud computing firms now are moving into the incipient business at the same time 5G suppliers hope to secure a position. But tower owners and some retailers might also expect they have a role as well. 

In the past, connectivity providers have often failed to compete with cloud computing suppliers or with independent data center providers, either. Verizon and AT&T. for example, already seem to be taking different roles in an effort to monetize existing assets for edge computing. 

Colocation and connections might be the role telcos eventually are forced to assume, their aspirations notwithstanding. Interoperability will be important for connectivity providers as they seek to serve global enterprises, allowing a single “throat to choke” offer for facilities scattered around the globe. 

At least so far, though, such capabilities seem more advanced at the level of colocation and hosting than actual computing as a service, at least in part because the current standards mostly apply to the ways different telcos operate their networks to support edge computing. 

The ETSI GS MEC 003 addresses the implementation of MEC applications as software-only entities that run on top of a virtualization infrastructure, which is located in or close to the network edge. A key new aspect of the Phase 2 version of this specification is the addition of MEC-in-NFV reference architecture, which defined how MEC-compliant edge deployments can be part of an overall NFV cloud architecture. 

ETSI aims for a unified edge compute architectural framework and reference platform, to be deployed across multiple markets in Europe and progressively extended to other operators and geographies to achieve global reach, ETSI says. 

The Framework and Reference Architecture standard includes the functional elements, the reference points between them, and a number of MEC services. In the updated version from 2019, a MEC variant was described with network functions virtualization (NFV) functional elements.

The Technical Requirements specification covers generic principles of MEC, such as NFV alignment and deployment independence. The 2018 update, Use Cases and Requirements, added application mobility to and from an external system, as well as 13 additional possible MEC use cases.

In 2019, the MEC ISG added 16 terms to the Terminology paper, including those often used in the ETSI MEC standards.

Mobile Edge Management has a two-part standard. The first, on system, host and platform management, defines the management protocol of the mobile edge system, hosts, and platforms. 

The second, on application lifecycle, rules, and requirements, outlines the application lifecycle management protocol in this standard. The document lists the rules and management requirements.

The User Equipment (UE) application interface standard details how to manage the application’s lifecycle on the connected device’s application interface.

The Bandwidth Management API standard primarily deals with bandwidth concerns when multiple devices use the same edge network. It focuses on application policy information, and how to address certain application program interface (API) scenarios that affect bandwidth usage and the network edge.

ETSI’s standard for the UE Identity API mainly focuses on a way to tag and track the user’s equipment in the network to enforce traffic rules. The standard for the Location API establishes guidelines for detecting the user’s device location information on the edge network.

In order to have a uniform Radio Network Information Service (RNIS) in MEC deployments, ETSI made the Radio Network Information API. It informs edge applications of the radio network’s condition to optimize network usage.

The Mobile Edge Platform Application Enablement document focuses on how the mobile platform functionality one (Mp1) reference point enables applications to communicate with the mobile edge system.

The General Principles for MEC Service APIs standard is a glossary of the RESTful API mobile edge service’s design principles, and highlights API guidelines and templates. In the 2019 update more RESTful API patterns were included.

The Support for Regulatory Requirements standard describes infrastructure to allow for Lawful Interception and Retained Data when implementing MEC into a larger network. This standard gives full support to these two practices’ regulatory requirements.

Is Edge Computing a Substitutute for Network Slicing?

Virtual private networks are not new in the core network. But network slicing, which allows 5G mobile networks to create end-to-end virtual private networks, is new. Up to this point, best effort has been the only possible quality of service level for mobile networks. 

Network slicing creates the ability to add quality of service, end to end, for mobile devices and networks. That will allow 5G service providers to create virtual networks, operating end to end, with defined network performance or features. 

Much will depend on how other trends, such as edge computing, play out. If guaranteed throughput or specified low latency are requirements, one way of satisfying those needs is to put local computing into place. 

In many proposed use cases--factory automation, IoT sensors for agriculture or connected car apps--edge computing is an alternative to network slices that run through the mobile core network. A high-bandwidth local network with local servers then becomes an alternative to a network slice. 

To be sure, a network slice might also be set up to connect a metro edge computing asset to “in the metro” sensors, assuring a minimum level of bandwidth and latency performance. But it might often be possible to use premises computing and local area networks to provide the required levels of bandwidth or latency performance. 

About 75 percent of service provider executives surveyed by Amdocs believe the internet of things, connected cars and smart homes represent early use cases for network slicing. 

About 20 percent of respondents also believe there will be network slicing use cases in industries such as mining, agriculture and health.

source: Amdocs

There are other alternatives to network slicing as well. Enterprises long have created their own VPNs for security purposes, purchased wholesale capacity or built and run their own wide-area optical networks. 

So the issue is whether mobile operators will want to supply enterprise customers with control of their own network slices, or whether some functional substitute for enterprise control can be created instead. In other words, is there a possible role for network slicing as the enabler of a new type of wholesale service?

Often other service providers are the customers, and that could be attractive. 

In principle, a network slice could be purchased by a mobile operator as an alternative to some other capacity arrangement, possibly with the upside of quality of service guarantees to the network edge. 

Perhaps that happens in a horizontal way, as often is the case, where functions are supplied. As 3GPP defines a slice, it is “an end to end logical communication network, within a Public Land Mobile Network (PLMN) and includes the Core Network (CN) Control Plane, User Plane Network Functions and 5G Access Network (AN).”

That allows the functions necessary to create a data connection between devices and servers. 

At least in principle, a private 5G network, operated by a single enterprise, on its own facilities, would offer full vertical control of the whole network. The analogy is a complete 5G network featuring only five or six cells. 

Some might see this as a threat to the mobile operator wide area business. Some tend to think even a vertical 5G private network is an extrapolation of the local area network concept, and so not a threat to WAN services. 

source: Amdocs

Streamed Apps Would Favor Edge Computing

Local computing on devices or local servers and remote computing using communications to reach servers have in the past been partial product substitutes. That will be true for edge computing as well. In fact, for new classes of applications, edge computing will provide the same business value as content delivery networks have provided for cloud apps generally: better performance and lower wide area network charges. 

source: Semantic Scholar

As latency performance and cost savings on bandwidth were the reason content delivery networks have had value, so edge computing will allow a reduction of north-south traffic. That might be especially true for use cases involving huge amounts of raw data to be processed and requiring artificial intelligence including real-time use cases. 

One development that could accentuate the tradeoffs is if apps could be streamed to devices as video and audio now are streamed. That would substitute remote servers and communications for local processing and storage on a device.

As always, the precise trade offs would depend on processing intensity and distance from the remote servers. Edge computing obviously would be helpful for processing intensive apps and use cases.

Cloud Infrastructure Revenue Grew 38% in 2019, Edge Impact Comes Next

AWS remains the biggest provider of cloud infrastructure services with $34.6 billion revenue. Microsoft generated  $18.1 billion, according to Canalys.

Global cloud infrastructure services spending grew 37.6 per cent to US $107.1 billion in 2019.

Google grew revenue the fastest, at 87.8 percent in 2019, albeit from a relatively low base, to $6.2 billion. 

Microsoft grew 64 percent to $18.1 billion, while Alibaba grew 64 percent as well,to $5.2 billion. AWS grew 36 percent to $34.6 billion. 

The “law of a few” or “winner take all” trend also seems to hold in the cloud infrastructure market. 

All four cloud kings  made their gains at the expense of "others", which lost collective market share of nearly five percent.

source: Canalys

In terms of edge computing, AWS seems to be moving most aggressively at the moment, launching three different edge computing services.  AWS Wavelength embeds AWS compute and storage services within telecommunications provider data centers at the edge of the 5G networks.

AWS Local Zone extends edge computing service by placing  AWS compute, storage, database, and other select services closer to large population, industry, and IT centers where no AWS Region exists today. 

AWS Local Zones are designed to run workloads that require single-digit millisecond latency, such as video rendering and graphics intensive, virtual desktop applications. Local Zones are intended for customers that do not want to operate their own on-premises or local data center.

Likewise, AWS Outposts puts AWS servers directly into an enterprise data center, creating yet another way AWS becomes a supplier of edge computing services. “AWS Outposts is designed for workloads that need to remain on-premises due to latency requirements, where customers want that workload to run seamlessly with the rest of their other workloads in AWS,” AWS says.  

AWS Outposts are fully managed and configurable compute and storage racks built with AWS-designed hardware that allow customers to run compute and storage on-premises, while seamlessly connecting to AWS’s broad array of services in the cloud.

Asia Tier 2 Hyperscale Data Center Growth

To the extent that hyperscale data centers now drive capacity requirements for the global wide area networks, it also stands to reason that new hyperscale data centers in “tier two” locations will drive capacity requirements. 

source: Analysys Mason

Who Wins at the Edge?

Dean Bubley

Edge computing almost certainly is going to create opportunities, but for whom? Alibaba and China Tower are working to create edge computing services, said Yang Yang, ShanghaiTech University School of Information Science and Technology co-director and professor.

But note something about that partnership: it is the app provider and the tower company that are working to create edge computing, not the connectivity provider. 

Amazon Web Services has partnered with several global telcos to create Wavelengths, an edge computing service using telco real estate. Initial partners include Verizon, Vodafone, KDDI and SKTelecom. But note there that that the edge computing service is supplied by AWS. The telco role is connectivity and rack space. 

Yang Yang

AT&T seems, at least initially, to favor a prioritized routing approach for enterprise data centers, essentially staying out of the colocation space. Another version of the AT&T service has Microsoft Azure supplying the edge computing, colocated inside enterprise data centers. 

The point is that multi-service edge computing might wind up being a smaller business for connectivity providers than had seemed possible a few years ago. 

John Ghirardelli

How much change will 5G bring to the rest of the internet ecosystem?  “Less than you think” or “almost everything” are the range of expectations for 5G expressed at a PTC’20 panel on 5G implications. Much hinges on which part of the ecosystem one examines. 

Mobile network physical infrastructure will change, but the business model could change even more. 5G will have the greatest and early impact on the telecom industry itself, argued Dean Bubley, Disruptive Analysis owner. New providers will emerge, including wholesale roles, private 5G operators and enterprise 5G networks as well, he says. 

Ramy Katrib

The physical layer changes might be least revolutionary. “5G is just another evolution of the ecosystem, using a wider range of frequencies, more fiber, bringing new entrants into the space and expanding the ecosystem,” said John Ghirardelli, American Tower Corporation director. 

On the other hand, there are many new technologies, including MIMO, network slicing and edge computing, said Yang,

Dense small cell networks with fiber extensively and deeply deployed are some other changes 5G will bring.

Gary Kim

Applications and business models might be where the greatest 5G changes happen. All prior mobile generations were built around human end users on phones. The 5G network is the first where machines, sensors and servers will outnumber phones, and where computers--not people--are the most numerous devices. 

That alone changes the business model from a consumer-driven mobile phone business to an enterprise-focused sensor connection business. In short, enterprise internet of things might drive new revenue, not consumer smartphone use. 

Also, 5G will differ from 2G, 3G or 4G, which served people using intelligent devices, said Yang. In comparison, most 5G device will be “stupid,” requiring computing support at various places in the network, autonomously on devices, someplace in the metro area or remotely. 

Which parts of the computing ecosystem benefit is the big question. It might be easy to suggest that application providers, device suppliers, edge computing as a service suppliers and others might benefit more than connectivity providers. 

But 5G could revolutionize scripted television production, because “a big pain point is connecting video production to the internet,” said Ramy Katrib, DigitalFilm Tree founder and CEO. “We compete with FedEx or physical media,” he said. 

And one key requirement is ensuring data recoverability even if connections are interrupted. “We assume there will be service interruptions,” Katrib said. So a key “killer app” is the assurance that all data is recoverable, even in the face of file transfer interruptions. 

Moving video files straight to editing suites will be revolutionary, Katrib argued. 

And patience might be required. “5G will be really popular by the time 6G arrives,” Yang quipped. “We are early days in terms of edge computing,” said Ghirardelli.

Verizon and AT&T Initially Take Different Approaches to Edge Computing

With the caveat that each firm might undertake other edge computing initiatives, Verizon and AT&T seem to be taking different approaches to the edge computing business. Verizon seems to have chosen to operate as a supplier of edge rack space and facilities. Verizon also is partnering with Amazon Web Services, which supplies the actual edge computing as a service.

AT&T seemed initially to prefer a version of enterprise edge computing, where AT&T routes traffic differentially based on enterprise priorities, but does not seem to require additional computing functions aside from that provided by the enterprise itself.

But a Microsoft Azure compatible version also seems to be available, where enterprises Microsoft Azure services in edge locations closer to customers and devices. In that scenario, it appears Azure computing services are available locally. 

source: AT&T

AT&T colocates routing gear at the enterprise data center and then routes traffic differentially. High-priority, mission-critical data using either Wi-Fi or mobile networks is processed locally and immediately sent back to the appropriate end-point within the customer’s private wireless network environment, rather than being processed remotely. 

It is possible, perhaps even likely that other edge initiatives also will be developed be each firm. 

Enterprise Workloads Now 79% Cloud

Public cloud spend is quickly becoming a significant new line item in information technology budgets, especially among larger companies, a survey sponsored by RightScale suggests. 

Among all respondents, 23 percent spend at least $2.4 million annually ($200,000 per month) on public cloud while 33 percent are spending at least $1.2 million per year ($100,000 per month). 

Among enterprises the spend is even higher, with 38 percent exceeding $2.4 million per year and half (50 percent) above $1.2 million per year. 

Small and medium businesses generally have fewer workloads overall and, as a result, smaller cloud bills (just over half spend under $120,000 per year). However, 11 percent of SMBs still exceed $1.2 million in annual spend, RightScale says. 

source: RightScale 

Enterprise respondents run 79 percent of workloads in cloud, with 38 percent of workloads in public cloud and 41 percent in private cloud. Workloads running in private cloud may include workloads running in existing virtualized environments or bare-metal environments that have been “cloudified,” says RightScale. 

Non-cloud computing comprises about 21 percent of respondent workloads. 

Small and mid-sized businesses report running 43 percent of workloads using public cloud and also run 35 percent of workloads on private cloud. Some 22 percent of workloads remains on non-cloud platforms. 

source: RightScale