|
Getting your Trinity Audio player ready...
|
Orange: From Months to Minutes – A Softwarized Network
Orange’s Journey: Jean-Louis Le Roux, EVP of International Networks at Orange, described how Orange is transforming its massive global network to be more flexible and cloud-like. With traffic doubling every few years and bandwidth needs becoming increasingly “on-demand,” Orange can no longer provision services in months – it needs to do so in minutes. To achieve this, Orange outlined a five pillar strategy
- Infrastructure Investment: Upgrading fiber/backbone for 30% annual traffic growth.
- Disaggregation: Separating hardware/software (DriveNets trials), segment routing, and network slicing.
- Telco Cloud: Decoupling services from physical network layers.
- Virtualization: Turning functions into software services (SD-WAN, CDN, etc.).
- Automation: Zero-touch orchestration and AI operations.
This approach is already yielding results – Orange has 200+ PoP sites running disaggregated equipment and even a core network trial with DriveNets software. By embracing a software-driven, orchestrated network, Orange expects to scale services faster (from months to minutes) while cutting costs and delivering new on-demand offerings.
KDDI: Embracing Open Disaggregation for the AI Era
KDDI’s Open Network: Kenji Kumaki, GM & Chief Architect at KDDI (one of Japan’s top carriers), shared why KDDI moved to an open, disaggregated network model. With 50+ million subscribers, KDDI is experiencing skyrocketing traffic – especially driven by AI applications, which is growing “much higher than we expected”. Traditional monolithic routers made it hard to scale (different proprietary hardware for core, edge, peering, each with limits and unique spares). KDDI saw disaggregation as the way to break these silos.
Two years ago, KDDI deployed a distributed disaggregated backbone router (DDBR) built on DriveNets Network Operating System software and white-box hardware. The results have been outstanding: zero outages or software bugs in two years of production, proving the solution’s stability. Kenji Kumaki highlighted key success factors for disaggregation – software stability, strong vendor partnership (DriveNets and Broadcom worked closely with KDDI), and the flexibility to scale out easily (disaggregated cluster can exceed 900 Tbps capacity, far beyond any single chassis). Encouraged by this success, KDDI announced just before MWC25 that they are extending disaggregation to the rest of their core backbone, and plan to transform the edge network by 2028.
In short, to handle the coming “AI era,” KDDI is betting on an open network design that is agile and can scale rapidly.
Cox: A Network for the Unexpected
Cox’s Vision: GS Sickand, VP of Wireless Engineering at Cox Communications, offered the perspective of a large cable operator. Cox (the third-largest cable MSO in the US) has decades of legacy infrastructure across residential broadband, mobile, and enterprise services. The challenge Cox is facing now is how to evolve its network to support new business models and on-demand services while keeping costs in check. Traditional long planning cycles no longer work. “We need to know the use case before designing… those days are over,” GS declared. In other words, you can’t design networks years in advance for specific apps, because new use cases that you cannot even imagine or plan for will pop up in weeks or months.
That’s why Cox’s strategy is to build a “programmable, data-driven network” that can be reconfigured on the fly for whatever comes next. GS called it a “network for the unexpected,” meaning the network should flexibly support any new application without lengthy redesigns. Achieving this requires converging their many types of networks (cable/DOCSIS, fiber PON, wireless, etc.) to a unified, automated one. Cox is looking to break down silos and manage everything end-to-end, with extensive automation so the network can scale up or down “without human intervention.” By 2034, as Cox aims to greatly expand its customer reach, this agile approach will ensure it can launch new services quickly and stay ahead of customer demands.
The takeaway: modernize and simplify the network now and create a simplified infrastructure that can adapt to whatever the future brings.
AT&T: The Best Converged Broadband Company
AT&T’s Convergence: Igal Elbaz, SVP Network CTO at AT&T, outlined AT&T’s bold plan to become “the best converged broadband company” – delivering mobile, home internet, and enterprise connectivity from a single integrated network. Customers want one provider for all their connectivity, so AT&T is unifying its nationwide 5G wireless network and its multi-gig fiber network rather than running them as separate silos.
Openness is a big theme in AT&T’s transformation plans. Igal noted AT&T’s embrace of technologies like Open RAN and a new 5G standalone core. AT&T was an early pioneer of disaggregation in its IP core network. “We started our open disaggregation years ago in the core backbone…carrying close to 840 petabytes a day,” Igal shared, referencing DriveNets-based core routing solution running on white-boxes. Over 80% of AT&T’s network traffic now runs through this disaggregated core powered by DriveNets software on Broadcom’s silicon– a strong validation of the approach. Given that success, AT&T is extending the same disaggregated architecture out to the network edge to converge all services there as well. In short, AT&T’s converged network strategy – fueled by open, software-based infrastructure – is delivering on both customer experience and cost efficiency at a massive scale.
Broadcom: Reinventing Economics with 10× Efficiency
Game-Changing Economics: Ram Velaga, SVP/GM Core Switching at Broadcom, brought a hardware perspective – showing how merchant silicon and open architectures are fundamentally changing network economics. He gave a striking example: not long ago, a high-end router chassis with 128×100G ports cost about $1,000,000 and drew ~10 kW of power. Today, an equivalent white-box built with Broadcom’s latest merchant chips can deliver the same 128×100G capacity for around $10–20k and under 1 kW of power. That’s roughly a 50× price reduction and 10× power efficiency gain – an almost unbelievable leap. These gains are thanks to Moore’s Law and the huge volumes driven by cloud data centers, which Broadcom has harnessed for telecom.
Ram noted that the rise of AI is pushing bandwidth needs even further (one AI server can consume hundreds of Gbps, and GPU clusters require terabits), meaning networks must scale like never before. Merchant silicon makes that scaling economically feasible. Broadcom’s role is providing the cutting-edge chips that “completely change the economics of this business,” as Ram put it, but he emphasized hardware alone isn’t enough – you need great software (like DriveNets Network Operating System) to tie it all together. He recounted how around 2015, AT&T challenged the industry to follow the cloud model on cost-per-bit. Broadcom, an ODM (UfiSpace), and a then-startup (DriveNets) teamed up with AT&T to build a fully disaggregated core router. Many were skeptical this radical approach would work, but today it’s proven – AT&T’s core runs on it, and “80%+” of their traffic goes through that architecture. Trailblazers like AT&T and KDDI took the first leaps so that now the rest of the industry can benefit from disaggregated economics.
The message: by embracing open hardware and software, operators can achieve cloud-level cost and power efficiency at telecom scale.
Simplifying Operations Through Automation
Unified Orchestration: A recurring theme in the panel discussion was the need to radically simplify network operations. After building these converged, disaggregated networks, how do you manage them end-to-end? The answer: break the old silos and orchestrate everything as one cloud. Jean-Louis (Orange) explained that Orange has built a “unique service orchestration” platform that manages everything from optical transport up to the mobile core on a common system. All network domains are controlled together with unified tools and automation. “If you don’t do this, you are dead,” he said bluntly, underscoring that tomorrow’s networks simply cannot be run in isolated silos. In other words, operators must operate across traditional boundaries (IP, optical, wireless, etc.) as one converged network, or they will fall behind in agility and cost.
GS (Cox) agreed and noted how operations must change in the era of cloudified networks. In the old days, if a single router or “box” failed, you replaced that hardware and restored service. In a disaggregated, virtualized network, services aren’t tied to one box – they run across many appliances and locations. This makes troubleshooting more complex unless you have the right tools. The goal is to reach a “zero-touch” state where 90%+ of incidents auto-resolve without human intervention.
The takeaway: unified orchestration and automation are not optional; they’re essential for managing converged networks at scale.
Automation and AI: Toward Zero-Touch Networks
DriveNets’ Automation Perspective: The panel also explored how automation and AI are key to running these new networks. DriveNets co-founder Hillel Kobrinsky highlighted that making the network “programmable” isn’t just about opening interfaces – it requires powerful automation so operators can actually harness that programmability. The goal isn’t to turn network engineers into software developers, but to augment them with automation tools so they can roll out new services much faster. One principle DriveNets is pursuing: minimize human touch. “In the new architecture… there is no use of CLI. People will not touch the network,” Hillel emphasized. Instead, an automation platform acts as a “programmable layer” on top of the disaggregated routers, translating high-level intents into network actions. This enables true zero-touch operations – the network can configure, heal, and optimize itself under software control for many routine tasks.
The discussion also touched on the growing role of AI in network operations. The DriveNets team demonstrated an AIOps capability where you can ask in plain English what’s wrong in the network, and an AI (like a ChatGPT for the network) diagnoses the root cause from hundreds of events. It’s an exciting glimpse of how AI can simplify troubleshooting in complex environments. Ultimately, highly automated, AI-assisted networks will let telcos move at “cloud speed” in deploying and managing services. And importantly, humans won’t be replaced – but those who leverage AI and automation will outperform those who do not. The vision is a network that runs itself as much as possible, allowing engineers to focus on innovation rather than manual maintenance.
New Services and Revenue: Monetizing the Transformed Network
Beyond Cost Cutting: A key audience question asked: with all these efficiencies (especially using DriveNets and disaggregation to cut costs), how will operators grow the top line? The panelists responded that a more agile network isn’t just about saving money – it’s about enabling new revenue through faster service innovation. GS (Cox) noted that nobody has a crystal ball for the next “killer app.” But if you build a network that can evolve quickly, you’ll be ready to support whatever new opportunity arises. Future use cases will arrive faster and with less warning, so the network must be flexible end-to-end (across cable, fiber, wireless) to accommodate them without months of re-engineering. In short, build it agile, and new revenue will follow.
Igal (AT&T) gave a concrete perspective: “At the end of the day, our network is our product,” he said. For AT&T, reducing time from idea to market is crucial. Disaggregated, software-driven networks not only save costs but also make the network more configurable and intelligent, allowing AT&T to create and deploy new services much faster. He cited examples like integrating security into the network itself (a dynamic defense service) or offering performance boosts like AT&T Turbo for wireless – innovations that are easier to deliver on a programmable network. The faster iteration means AT&T can experiment with new features in months instead of years.
Ram (Broadcom) zoomed out to the industry level, observing that we’re entering an era of massively distributed computing where “the network is the computer”. In cloud computing, the network played a supporting role, but in distributed AI (with workloads spread across many data centers and edge locations), the network becomes absolutely central. This presents huge opportunities for service providers. Hyperscalers are now hunting for sites with power and fiber to run distributed AI workloads – and telcos, with their extensive fiber networks, are in a prime position to host these new services. A transformed network can do far more than carry bits – it becomes a platform for futuristic services that customers will pay for.
Conclusion: The Network Transformation Journey
The MWC25 panel made it clear that network transformation is no longer a theory – it’s happening now across the industry. Telecom giants and innovators alike are aligned on a shared vision of an open, converged, and automated network that can meet tomorrow’s demands. From Orange and KDDI to AT&T and Cox, operators are re-architecting their networks for agility and efficiency, with DriveNets emerging as a key enabler of this industry-wide movement. These insights underscore that transforming the network pays dividends not just in cost savings, but in speed, innovation, and new revenue opportunities.
Related content for Service Provider Routing and Network Infrastructure
eGuide
Transforming Service Provider Networks