Welcome to “Complex Made Clear” – where we take the intimidating, untangle the technical, and make the complicated feel conquerable. In this series, we break down complex topics into digestible, easy-to-understand pieces without losing their essential meaning. Whether you’re a seasoned professional looking to grasp new concepts or someone just starting their journey, each article serves as your friendly guide through the maze of modern technology and business concepts. No jargon without explanation, no assumptions about prior knowledge – just clear, straightforward explanations that help you have those “oh, now I get it!” moments. Let’s make the complex clear, together.
From terabit-scale 5G deployments to industry-wide infrastructure standards: how leading operators are actually implementing next-generation network automation. After two posts discussing problems and solutions, you’re probably thinking: “This all sounds great in theory, but does it actually work in the real world?” I get it. The networking industry has been burned by overpromised automation solutions before. But here’s the thing – some operators have moved beyond proof-of-concepts and pilots. They’re running these open source solutions in production, at scale, with measurable business impact. Let me show you what’s actually working out there.
Verizon’s Bold Bet: Internalizing SDN Control
When most telecommunications operators were evaluating vendor-packaged SDN solutions, Verizon made a decision that surprised many in the industry: they chose to internalize OpenDaylight, creating their own custom distribution and building internal expertise to manage it. Three years later, this strategy has proven to be one of the most successful large-scale SDN deployments in the industry.
The numbers tell the story. Verizon now runs OpenDaylight as their “foundational and directional SDN controller” across production networks, handling both NETCONF-based provisioning for their Intelligent Edge Network and OpenFlow-based packet flow management for their native OpenFlow infrastructure. But the real value isn’t in the technology itself—it’s in what the technology enabled.
Here’s what makes this interesting: instead of taking OpenDaylight as-is, Verizon decided to internalize the entire stack. They created their own testing, packaging, optimization, and support organization. Why? They were frustrated with vendor options that felt inflexible and created lock-in concerns.
The results? They’re using it in two critical production scenarios:
- NETCONF-based provisioning for their Intelligent Edge Network (iEN), replacing costly vendor Element Management Systems
- OpenFlow controller managing end-to-end packet flows across native OpenFlow infrastructure
The value proposition is compelling: vendor-agnostic abstraction that simplifies integration with upstream OSS/BSS systems, elimination of expensive vendor EMS licenses, and the ability to leverage community innovation without vendor lock-in.
But here’s the honest truth – it wasn’t easy. Verizon initially struggled to find OpenDaylight expertise and had to invest heavily in building internal capabilities. They’re also working on horizontal scaling challenges as their deployments grow.
The lesson? When you’re serious about avoiding vendor lock-in and gaining control over your network automation destiny, the investment in open source expertise pays off. But “serious” means really serious – not just a side project.
Orange’s ONAP Production Reality Check
Orange provides one of the most detailed public case studies of ONAP in production. Their deployment in Orange Egypt focused on automating IP/MPLS backbone operations – exactly the kind of complex, mission-critical infrastructure where automation failures hurt.
Orange’s approach is impressive because of its pragmatic component selection. They didn’t try to deploy the entire ONAP platform at once. Instead, they focused on specific components (SO, A&AI, SDN-C, CDS) that solved their immediate automation challenges.
Orange found these components “mature and flexible enough for production use,” highlighting the value of A&AI for modeling complex network relationships and the ease of integration for components like SO and CDS.
The business impact? Orange positioned this as central to their “Lead The Future” strategy, viewing ONAP as essential for achieving network automation goals while avoiding vendor lock-in.
Samsung’s 5G UPF Performance Breakthrough
Now let’s talk about something that really gets network engineers excited: Samsung achieving over 1 Terabit per second throughput for their 5G User Plane Function using FD.io VPP and Intel hardware.
This isn’t just impressive marketing numbers – it’s a fundamental validation that software-based network functions can compete with dedicated hardware appliances at carrier scale.
Here’s what makes this particularly clever: Samsung leveraged Intel E810 NIC’s Dynamic Device Personalization (DDP) feature to offload GTP-U header parsing to hardware. This means the NIC can examine inner packet headers and distribute traffic across CPU cores intelligently, freeing up processing power that was previously used for packet distribution.
The technical elegance is beautiful, but the business implications are profound. Software-based 5G core functions can now deliver carrier-grade performance on commodity hardware, fundamentally changing the economics of 5G infrastructure.
Mavenir’s Energy Efficiency Play
While Samsung focused on raw performance, Mavenir took a different approach with their VPP-based UPF: energy efficiency through SmartNIC offload.
Their results are striking – achieving the same 524 Gbps throughput while using approximately 50% less CPU compute power through NVIDIA SmartNIC acceleration. In an era where energy costs and sustainability matter more than ever, this kind of efficiency gain is game-changing.
This isn’t just about performance – it’s about Total Cost of Ownership (TCO). Lower CPU utilization means:
- Reduced server hardware requirements
- Lower power consumption
- Better heat dissipation
- Higher infrastructure utilization
- Improved sustainability metrics
Mavenir’s approach demonstrates that the future of high-performance networking isn’t just about raw software optimization – it’s about intelligent hardware-software co-design.
Bell Canada’s VNF Onboarding
Sometimes the most impressive results aren’t the flashiest ones. Bell Canada’s ONAP deployment focused on something unglamorous but critical: automating VNF onboarding and lifecycle management.
Their reported outcome? “Significant savings in recurring manual work” and dramatically accelerated onboarding times for new network services. While they didn’t publish specific numbers, anyone who’s struggled with VNF onboarding knows how painful the traditional process can be.
The transformation from a 3-month manual process to a 2-week automated workflow represents the kind of operational efficiency gain that directly impacts business agility.
Deutsche Telekom’s O-RAN SMO Success
Deutsche Telekom deployed ONAP as the vendor-independent Service Management and Orchestration (SMO) framework in their multi-vendor O-RAN trial network in Neubrandenburg, Germany. This was noted as the first live network integration of multi-vendor Open RAN with massive MIMO in Europe.
What makes this significant isn’t just the technical achievement – it’s the strategic implications. By using ONAP to manage the complete lifecycle of O-RAN components, DT demonstrated that open source automation can handle the complexity of next-generation RAN architectures.
The proof point here is about vendor diversity and interoperability. When you can orchestrate equipment from four different vendors in a live network trial, you’ve demonstrated that open source automation can break down traditional integration barriers.
Nokia’s Nephio Reality Check
Nokia Bell Labs conducted one of the most thorough real-world evaluations of Nephio, using Release 2 to deploy a Nokia 5G core network function suite on AWS infrastructure. This wasn’t a synthetic demo – they mirrored their commercial offering in a realistic cloud environment.
The results were mixed, and honestly, that makes this case study more valuable than pure success stories.
What worked well:
- Nephio’s core principles (Configuration-as-Data, intent-driven automation, Kubernetes control plane) proved sound
- The GitOps workflow enabled version control and audit trails for complex network configurations
- Integration with AWS infrastructure worked as designed
What needed improvement:
- Significant effort required to adapt existing Nokia KRM resources into Nephio-compatible blueprints
- Difficulties accurately describing complex network topologies with existing Nephio constructs
- Need for additional features to enable true one-click lifecycle management
What I appreciate about Nokia’s transparency is that it provides realistic expectations. Nephio’s architectural approach is sound, but real-world adoption requires investment in tooling, processes, and expertise. The trial successfully validated the concept while identifying concrete areas for improvement.
Spark New Zealand’s 5G Preparation
Spark New Zealand, partnering with Infosys, implemented ONAP in six months to establish an automation suite supporting future 5G use cases, specifically targeting network slicing and closed-loop automation.
Six months from start to production automation capability is impressive, especially for 5G-ready infrastructure. This timeline suggests that with proper partner support and focused scope, ONAP deployment can be much faster than early adopters experienced.
The key success factor seems to be clear scope definition – focusing on specific 5G automation capabilities rather than trying to automate everything at once.
China Mobile’s Multi-Use Case Validation
China Mobile shared their experience deploying ONAP blueprints for both vCPE (virtual Customer Premises Equipment) and CCVPN (Cross-Carrier VPN) use cases, demonstrating ONAP’s versatility across different service types.
What’s particularly interesting is their multi-use case approach. Instead of betting everything on one service type, they validated ONAP’s capabilities across different scenarios, building confidence in the platform’s general applicability.
The Performance Revolution: Quantified Results
Let me put some specific numbers on the table from various FD.io VPP deployments:
Samsung 5G UPF: 1+ Tbps (with Intel 4th Gen Xeon + E810 NICs) Mavenir UPF: 524 Gbps (with 50% CPU reduction via SmartNIC offload) Generic VPP Benchmarks: Multi-hundred Gbps on commodity hardware
These aren’t laboratory curiosities – they’re production-ready implementations that fundamentally change the economics of software-based networking.
Common Success Patterns
Looking across these case studies, several patterns emerge:
- Pragmatic Component Selection Successful operators don’t try to deploy entire platforms at once. They identify specific automation challenges and select the right combination of LFN components to solve them.
- Investment in Expertise Whether it’s Verizon building internal OpenDaylight expertise or operators partnering with integrators like Infosys, success requires serious investment in understanding these technologies.
- Focus on Business Outcomes The most successful deployments tie technical capabilities directly to business metrics: reduced OpEx, faster service deployment, improved reliability, or vendor independence.
- Iterative Approach Nobody achieved their end state in one big bang. Successful operators start with focused use cases, learn, and gradually expand scope.
The Hard-Won Lessons
These case studies also reveal some uncomfortable truths:
Open Source Doesn’t Mean Free Every successful deployment required significant investment in expertise, integration, testing, and ongoing maintenance. The software might be open source, but the expertise isn’t free.
Customization is Inevitable From Verizon’s custom OpenDaylight distribution to Nokia’s blueprint adaptation challenges, every serious deployment requires some level of customization.
Standards Alignment Takes Time While these projects aim for standards compliance, real-world integration often reveals gaps that require workarounds or community contributions.
Performance Requires Hardware Awareness The most impressive performance results combine software optimization with hardware acceleration. Pure software approaches have limits.
What’s Coming Next
Based on these real-world experiences, I see several trends accelerating:
Hardware-Software Co-Design: The Samsung and Mavenir examples show that peak performance requires intelligent use of hardware acceleration.
AI-Driven Operations: Closed-loop automation is evolving toward AI-driven predictive operations that can prevent issues before they impact services.
Edge-Scale Automation: Projects like EMCO and Nephio are tackling the challenge of managing automation across thousands of edge locations.
Intent-Based Everything: The progression from imperative scripts to declarative configurations to intent-based automation is accelerating.
The Bottom Line
These aren’t just technology success stories – they’re business transformation examples. Operators using these approaches are achieving:
- 50% reduction in CPU requirements (Mavenir)
- Significant savings in recurring manual work (Bell Canada)
- Six-month implementation timelines (Spark New Zealand)
- 1+ Tbps software performance (Samsung)
- Vendor lock-in elimination (Verizon)
The networking industry is at an inflection point. The operators investing in open source automation capabilities today are building competitive advantages that will matter for years to come.
Those still debating whether this stuff works in the real world? The evidence is in. The question isn’t whether open source network automation works – it’s whether operators are going to be part of the transformation or left behind by it.
Feaured Image by Pete Linforth from Pixabay