Nvidia takes 5G to the edge with help from Ericsson and Red Hat

Graphics chip maker Nvidia has unveiled its EGX Edge Supercomputing Platform that is designed to boost 5G, IoT and AI processing at the edge of the network

Nvidia has long been the market leader in GPUs (graphics processing units), which has enabled it to get a strong position in supercomputing, where the parallel processing qualities of GPUs come in especially handy. This EGX initiative seems to be Nvidia’s attempt to translate that position from datacentres to the edge computing.

“We’ve entered a new era, where billions of always-on IoT sensors will be connected by 5G and processed by AI,” said Jensen Huang, Nvidia CEO. “Its foundation requires a new class of highly secure, networked computers operated with ease from far away. We’ve created the Nvidia EGX Edge Supercomputing Platform for this world, where computing moves beyond personal and beyond the cloud to operate at planetary scale.”

There seems to be a fair bit of support for this new platform, with a bunch of companies and even a couple of US cities saying they’re already involved. “Samsung has been an early adopter of both GPU computing and AI from the beginning,” said Charlie Bae, EVP of foundry sales and marketing at Samsung Electronics. “NVIDIA’s EGX platform helps us to extend these manufacturing and design applications smoothly onto our factory floors.”

“At Walmart, we’re using AI to define the future of retail and re-think how technology can further enhance how we operate our stores,” said Mike Hanrahan, CEO of Walmart Intelligent Retail Lab. “With NVIDIA’s EGX edge computing platform, Walmart’s Intelligent Retail Lab is able to bring real-time AI compute to our store, automate processes and free up our associates to create a better and more convenient shopping experience for our customers.”

On the mobile side Ericsson is getting involved to build virtualized 5G RANs on EGX. As you would expect the reason is all about being able to introduce new functions and services more easily and flexibly. More specifically Ericsson hopes the platform will make virtualizing the complete RAN solution cheaper and easier.

“5G is set to turbocharge the intelligent edge revolution,” said Huang. “Fusing 5G, supercomputing, and AI has enabled us to create a revolutionary communications platform supporting, someday, trillions of always-on, AI-enabled smart devices. Combining our world-leading capabilities, Nvidia and Ericsson are helping to invent this exciting future.”

On the software side a key partner for all this virtualized 5G fun will be Red Hat, which is getting its OpenShift Kubernetes container platform involved. It will combine with Nvidia’s own Aerial software developer kit to help operators to make the kind of software-defined RAN tech that can run on EGX.

“The industry is ramping 5G and the ‘smart everything’ revolution is beginning,” said Huang. “Billions of sensors and devices will be sprinkled all over the world enabling new applications and services. We’re working with Red Hat to build a cloud-native, massively scalable, high-performance GPU computing infrastructure for this new 5G world. Powered by the Nvidia EGX Edge Supercomputing Platform, a new wave of applications will emerge, just as with the smartphone revolution.”

Things seemed to have gone a bit quiet on the virtualization front, with NFV, SDN, etc having apparently entered the trough of disillusionment. Nvidia is a substantial cloud player these days, however, and judging by the level of support this new initiative has, EGX could a key factor in moving the telecoms cloud onto the slope of enlightenment.

Huawei pledges $1.5 billion to its new developer program

Huawei has announced that it will invest $1.5 billion in the next five years to boost its developer ecosystem for the Kunpeng and Ascend computing platforms.

SDKs were also released at the same event when its Developer Program 2.0 was unveiled.The announcement was made at the 2019 version of the Chinese vendor’s annual Huawei Connect event in Shanghai. According to Patrick Zhang, CTO of Cloud & AI Products & Services at Huawei, the new program will cover five key areas:

  • Building an open computing industry ecosystem based on Kunpeng + Ascend computing processors
  • Establishing an all-round enablement system
  • Promoting the development of industry standards, specifications, demonstration sites, and technical certification system
  • Building industry-specific application ecosystems and region-specific industry ecosystems
  • Sharing Kunpeng and Ascend computing power, making it available to every developer

The focus areas are related to cloud computing and artificial intelligence. The applications and services the ecosystem aims to support are for server level, either in the centralized cloud or on the edge. To enable the ecosystem development, Huawei also published Kunpeng Developer Kit and ModelArts 2.0 AI development platform.

Despite that x86 architecture is still dominating the server market, ARM has worked to break the monopoly, and Huawei is one of ARM’s leading licensees. Earlier this year Huawei released Kunpeng 920, its CPU based on ARMv8 design. Huawei aims to expand its share in the server market with Kunpeng’s superior computing power claimed by Huawei, most likely starting from the market in China.

But Huawei’s ambitions go way beyond moving more boxes. Its cloud service has been promoted for its strong AI capability, supported by the Ascend AI chips. The Ascend 910, the latest version, was released in August, which the company claimed is the world’s most powerful AI processor.

By enriching its ecosystems, Huawei hopes it will be able to deliver a full suite of solutions, including supporting digital transformation undertake by increasing numbers of telecom operators.

This is the second iteration of Huawei’s Developer Program. The Developer Program 1.0 was launched in 2015.

Should operators try to own the edge?

At the Edge Computing Congress 2019 in London, the keynotes and panel discussions focused on the unique opportunity for operators to own the edge if they want to.

Edge computing refers to distributed datacenters that place reduce the physical distance between the cloud and the edge of the network – i.e. the RAN. The main point of this is to reduce the lag from interacting with the cloud in real time and to allow the kind of low-latency communication services that promise to be the most novel new feature of 5G. Edge computing is also expected to help with things like bandwidth flexibility for IoT, cloud security and data localisation.

The presentations were opened by Julian Bright of analyst firm Ovum, who warned that 5G probably needs the edge more than the edge needs 5G and set the tone for the rest of the morning by asking who will own it. Bright also raised the issue of interoperability and noted that the definition of a common framework for edge computing is some way from being determined.

As is often the way, most of the talking points came from a panel (pictured) that aimed to explore definitions of edge computing, what the point of it is, and the business cases for investing in it. It was agreed that the edge is not a discrete, standalone thing, but rather an extension of the cloud. That said, by definition it requires its own separate physical infrastructure, which has to be built and owned by someone.

This is presents a unique opportunity for operators, for whom distributed infrastructure is a core competence. They also own, or at least have access to, a lot of remote locations, so they have a head start over cloud specialists and IT companies. Edge computing was said to be the perfect example of the convergence of networks and IT, which raises the question of which of those worlds will define and own it.

A key issue for edge computing concerns interoperability. As an extension of the public cloud it needs to be usable by all stakeholders. One way to ensure this is standardisation, something the telecoms world is very familiar with. Standardisation typically takes a long time, however, and the panel warned that operators are likely to lose their advantages in this space if they allow themselves to be bogged down by it.

There are also cultural dynamics involved. The IT world typically moves faster and is less risk-averse than the networking world. While telcos are used to significant infrastructure capex, this is typically in areas where there is proven demand and ROI. Heavy investment in edge computing will require more of a ‘build it and they will come’ strategic philosophy.

This observation led to a discussion of the chicken-and-egg dilemma that comes with the prospect of investing in a new technological platform without a mature business case to go with it. As we saw with historical attempts to break the duopoly in smartphone operating systems, it’s hard to get customers for your platform without a strong app ecosystem, but developers are reluctant to embrace any platform that doesn’t have a large and enduring user base.

There was unanimity among the panel that ownership of the edge is there for operators to take if they want it, but they need to move fast. If they do they will need to accept risk in the form of capex without the guaranteed ROI they’re used to and they will also need to seed the app ecosystem in ways they have historically avoided. For operators the edge is about new revenue opportunities rather than efficiencies and their approach to it needs to reflect that.

After the panel there was a keynote from Mahadev Satyanarayanan, Professor of Computer Science at Carnegie Mellon University, who further explored the value propositions of edge computing. He stressed that the deeper the use of the edge, the more of a premium can be charged for the resulting service. A real-world example of such a service can be seen in the video below of a project Satyanarayanan oversaw, using vehicle cameras to enable crowdsourced traffic information without the driver needing to be actively involved.

Edge computing: will 5G make it a reality?

Mobile edge computing, latterly expanded to multi-access edge computing and conveniently abbreviated MEC throughout, has been extensively promoted in the telecom arena for some time, though its actual broad implementation has been so far elusive. The ascendancy of 5G will change that. For one thing, the technology properties of 5G will facilitate the rollout of edge computing; for another, most of the leading 5G use cases will also demand more computing at the edge.

Here we are sharing the opening section of this Telecoms.com Intelligence special briefing to look at the status of mobile edge computing and considers how stakeholders can best address the opportunities.

The full version of the report is available for free to download here

Introduction: why it is a good time to talk about edge computing for 5G

Edge computing on mobile networks is by no means a new topic. One of the early demonstrations took place at Mobile World Congress (MWC) 2013, when Nokia (then Nokia Siemens Networks, NSN) introduced its base station-based technology Radio Applications Cloud Server (RACS), which supported real-time data processing of the radio and network information from the base station.

In 2014, the standardisation body ETSI established a working group, the Industry Specification Group (ISG) on MEC, which initially referred to Mobile Edge Computing but was expanded to cover Multi-access Edge Computing in March 2017. The group’s new remit then, by definition, would cover different access technologies. The founding members included both vendors (Huawei, Nokia, IBM, Intel) and operators (NTT CoCoMo, Vodafone). So far, the group has completed Phase 1 specifications and is in the middle of Phase 2 specifications.

What may seem disappointing to the edge computing enthusiasts is that no serious commercial rollout of MEC on a big scale has yet happened, although there have been plenty of proofs of concept (PoCs). As a matter of fact, part of the ETSI MEC group’s mission is to develop specifications for and provide support to PoC cases.

The industry has generally recognised that edge computing can deliver multiple benefits. Pushing computing to the edge will allow mobile operators to support new services for customers, especially applications reliant on cloud computing and demanding low latency. It will also help them to cut spending on backhaul and avoid congestion in the core.

There are also non-technical reasons why edge computing is desirable. Data sovereignty requirements in many countries mean data storage and processing must happen within national borders. Meanwhile, some companies are concerned about data leaks and demanding that information be stored on their own premises.

The market has warmed to edge computing thanks to the accelerated deployment of 5G, largely because the edge and 5G have a symbiotic relationship. While special-purpose mobile edge computing can already be done in 4G networks, 5G is the first mobile technology that can support it at full scale. If edge computing is to realise its potential for mobile stakeholders and adjacent industries, the industry and its partners will need to answer these critical questions:

  • Who should invest in, or “own,” the edge?
  • What will be the dominant dynamics between 5G and edge computing?
  • What type of edge computing service is most likely to be broadly adopted in the 5G environment?
  • What will be the biggest impediments to the broad adoption of edge computing in 5G?

The rest of this briefing includes sections on:

  • The whys and whats: industry sentiment and expert views on edge computing
  • Current state of play: who’s doing what?
  • Where are we going from here? Edge computing in the 5G era
  • Interview with Julian Bright, senior analyst for intelligent networks, Ovum

To access the full briefing please click here

Edge computing: will 5G make it a reality?

Mobile edge computing, latterly expanded to multi-access edge computing and conveniently abbreviated MEC throughout, has been extensively promoted in the telecom arena for some time, though its actual broad implementation has been so far elusive. The ascendancy of 5G will change that.

This Telecoms.com Monthly Briefing, including input from telecom practitioners as well as leading industry analysts, looks at the status of mobile edge computing and considers how stakeholders can best address the opportunities.

Please complete the short form below to access this briefing.

AT&T + HPE = edgy TLC

AT&T has announced a new partnership with HPE to drive the benefits of edge computing in enterprise services.

The duo has agreed a go-to-market strategy to accelerate business adoption of edge connections and edge computing, seen by some as one of the most interesting areas of the up-coming 5G economy.

“AT&T’s software-defined network, including our 5G network, combined with HPE’s intelligent edge infrastructure can give businesses a flexible tool to better analyze data and process low-latency, high-bandwidth applications,” said Mo Katibeh, CMO of AT&T Business. “Bringing compute power closer to our network helps businesses push the boundaries of what is possible and create innovative new solutions.”

“HPE believes that the enterprise of the future will need to be edge-centric, cloud-enabled and data-driven to turn all of its data into action and value,” said Jim Jackson, CMO of HPE. “Our go-to-market alliance with AT&T, using HPE Edgeline Converged Edge Systems, will help deliver AT&T MEC services at scale to help our customers more quickly convert data into actionable intelligence, enabling unique digital experiences and smarter operations.”

There are of course many benefits to edge computing, though one of the areas AT&T will be hoping to address through this tie-up will be the security concerns which will emerge. This looks like it could be one of the key marketing plugs of the AT&T proposition, as its Multi-access Edge Compute (MEC) Services will hope to drive the benefits of mobility to enterprise customers.

From HPE’s perspective, the team will be contributing on the low-latency side of the 5G euphoria. HPE suggests its Edgeline Converged Edge Systems could help create use cases where applications can reside on premises for lower latency processing.

It might not be as ‘sexy’ as plugging ridiculous download speeds, but the greatest benefits of 5G to the telcos would appear to be diversification as opposed to increased squeeze on the wallets of consumers. With more data being created each day, the edge will become increasingly important to activate products, services and business models in a faster and more operationally efficient manner. Enterprise organizations will largely be unaware of how to reap the greatest benefits, a pleasant niche the telcos could certainly profit from.

The edge takes centre stage (again)

At Mobile World Congress in Barcelona, the edge was one of the most prominent topics and the same can be said in Denver at Light Reading’s Big 5G Event.

The edge itself is a relatively easy concept to understand, but in reality, the deployment and evolution are much more complicated facets. It is usually the case that the biggest potential or most promising developments are the most difficult to achieve, but it does seem there is resistance continued resistance to change.

“I don’t think we are doing 5G right, we are too obsessed by being faster,” said Telus CTO Ibrahim Gedeon.

This is a message which is becoming increasingly frequent, and while the frequency is encouraging, an some point the walk is going to have to follow the talk. At this event, Telus, Equinix, AT&T, T-Mobile US and Verizon have all preached the benefits of moving beyond a ‘bigger, faster, meaner’ business model which is focused on speeds, but soon enough some might assume these are simply statements to appease the masses.

In terms of the future of the telco world, those telcos who maintain the status quo, focusing on speeds, will march down the path of commoditisation. Those who challenge themselves, taking on the incredibly difficult challenge of evolving the business, will find additional value and redefine themselves as a ‘Digital Service Provider’. And the edge is a very important component of that.

“Does 5G need edge, yes,” said Jim Poole, VP of Ecosystem Business Development at Equinix. “But does edge need 5G, no.”

This is a tricky position and a conundrum often faced in the telco and technology world. For some, the path forward does often look like a solution in search of a problem. It might not sound like the most attractive path, especially when there is the simplistic and more instantly gratifying pursuit of increased speeds.

It all depends on what the telco actually wants. As Mitch Wagner of Light Reading pointed out, there is nothing fundamentally wrong with the utility business model, these are companies which do make money and are often much safer bets, but this is not what the majority of telcos want in the future.

The answer is relatively plain if utilitisation is to be avoided; add more value to the ecosystem. This is where the likes of Google, Amazon and Microsoft have been successful. They have embraced the idea of a solution looking for a problem, demonstrated the value in disrupting the status quo and proved to the world new is better. For the telcos, this means embracing the unknown.

The edge is that unknown for the moment. Yes, the edge exists in the 4G world, but there is much more value in the 5G era. It is a case of putting the cart before the horse, but as it was pointed out today, creating a proposition which is defined by the edge offers a new dynamic, encourages the imaginations of a new type of developer and creates new services. Telcos can be fundamental in delivering the foundations of these new businesses, products and services, but they need to disrupt themselves and create a new organization.

This again is a statement which is not new. This week and countless other conferences have seen executives preach the promise of evolution, but sooner or later the telcos are going to have to replace talking with walking.

The time of rhetoric is coming to an end, we need to see progress in these transformation initiatives.

The private power of the edge

One of conundrums which has been quietly emerging over the last couple of months concerns how to maintain privacy when attempting to improve customer experience, but the power of the edge might save the day.

If telcos want to be able to improve customer experience, data needs to be collected and analysed. This might sound like a very obvious statement to make, but the growing privacy movement across the world, and the potential of new regulatory restraints, might make this more difficult.

This is where the edge could play a more significant role. One of the more prominent discussions from Mobile World Congress in Barcelona this year was the role of the edge, and it does appear this conversation has continued through to Light Reading’s Big 5G Event in Denver.

Some might say artificial intelligence and data analytics are solutions looking for a problem, but in this instance, there is a very real issue to address. Improving customer experience though analytics will only be successful if implemented quickly, some might suggest in real-time, therefore the models used to improve performance should be hosted on the edge. This is an example of where the latency business model can directly impact operations.

It also addresses another few issues, firstly, the cost of sending data back to a central data centre. As it was pointed out today, telcos cannot afford to send all customer data back to be analysed today, it is simply an unreasonable quantity, therefore the more insight which can be actioned on the edge, with only the genuinely important insight being sent back to train models, the more palatable customer experience management becomes.

Secondly, the privacy issue is partly addressed. The more which is actioned on the edge, as close to the customer as possible, the lesser the concerns of the privacy advocates. Yes, data is still being collected, analysed and (potentially) actioned upon, but as soon as the insight is realised the sooner it can be deleted.

There are still sceptics when it comes to the edge, the latency business case, artificial intelligence and data analytics, but slowly more cases are starting to emerge to add credibility.