Part-2: Unleashing Network Potentials: Current State and Future Possibilities with AI/ML

By Vinay Tripathi with Ajay Lotan Thakur


In the dynamic realm of networking, AI/ML has emerged as a transformative force to reshape the networking world by making it more secure, reliable, efficient and optimized. In this blog we will dive into characteristics, possibilities, use cases and challenges of AI/ML in the networking.

About AI and ML

Definitions of AL/ML

AI and ML are often used interchangeably, but there are some key differences between the two.

  • AI is the ability of machines to perform tasks that would normally require human intelligence, such as understanding natural language, recognizing objects, and making decisions.
  • ML is a subfield of AI that allows machines to learn from data and improve their performance over time.
  • DL = Uses neural networks for complex structured models and greater insights.

Types of AI/ML

AI/ML encompass a wide range of techniques and algorithms that can be used to solve a variety of problems. In the context of networks, AI/ML technologies can be broadly categorized into the following types:

Key Points:

  • AI/ML taxonomy is continuously evolving due to industry growth and various methodologies and algorithms.
  • The choice of AI/ML algorithm significantly influences business outcomes, including training time, prediction accuracy, and resource usage.
  • The selection of algorithms depends on the type and volume of available data for a specific use case.

Popular ML Types:

  • Supervised/Unsupervised: When available data is simple or significant pre-processing has resulted in high data quality:
  • Neural Networks and Deep Learning: When you have substantial amounts of unstructured/structured data or unclear features these may offer superior accuracy over Classical ML methods
  • AutoML: When you need to streamline machine learning model development, especially with limited expertise, time, or resources.
  • NLP: When tasks involve text or language data and require automation, understanding, or generation of natural language content.
  • Reinforcement learning: Suitable when you need to train agents to make sequential decisions in dynamic environments, optimizing for long-term rewards, and when there is a need for autonomous decision-making, such as in robotics, game playing, or autonomous systems.

    Figure-1: Hierarchy of AI, ML and DL

Applications of AI/ML

AI and ML technologies provide a diverse array of applications in networks, encompassing security, engineering, capacity planning, and operations. These technologies have the capability to augment network security, optimize network design and performance, forecast traffic demand, and automate network tasks. This leads to enhanced efficiency, reliability, and overall network performance. Here are some specific examples:

Network Security

  • Intrusion Detection System (IDS): AI-powered IDS can detect and respond to cyberattacks in real-time, providing a more robust defense against threats.
  • Thread Detection and Prevention (TDP): AI can analyze network traffic to identify and prevent threats before they can cause damage.
  • Anomaly Detection: AI can detect deviations from normal network behavior, indicating potential security incidents.

Network Engineering

  • Quality of Service (QoS): AI can optimize network resources to ensure consistent and reliable performance for critical applications.
  • Routing and Traffic Management: AI can optimize routing decisions and manage traffic flow to avoid congestion and improve network performance.
  • Optimized Traffic Flow: AI can analyze traffic patterns and make real-time adjustments to optimize traffic flow, reducing latency and improving overall network performance.
  • Load Balancing: AI can distribute traffic across multiple servers or network links to balance the load and prevent bottlenecks.

Network Capacity Planning

  • Improved Capacity Forecasting: AI can analyze historical data and predict future traffic demand, enabling network operators to plan for future capacity needs.
  • Efficient Uses of Resources: AI can identify and allocate network resources more efficiently, reducing costs and improving network performance.

Network Maintenance, Troubleshooting, Operations and Monitoring

  • Real-time Monitoring: AI can continuously monitor network performance and identify potential issues before they cause outages or disruptions.
  • Quicker Resolutions of Vendor/Hardware Issues: AI can diagnose and resolve vendor and hardware issues more quickly, minimizing downtime.
  • Faster Root Cause Analysis: AI can analyze large amounts of data to identify the root cause of network issues, enabling faster resolution.
  • Quick Mitigations of Network Issues: AI can automatically implement mitigations for network issues, reducing the impact on users and applications.

AI/ML Based Network in Action

The seamless integration of AI/ML components at various levels of the network (edge, core, management, etc.) enhances its reliability, efficiency, and security by optimizing performance and safeguarding against vulnerabilities.

The diagram illustrates a practical application of AI/ML within one of the extensive networks.

Figure-2: AI/ML in action in a cloud network

Trends in AI/ML

AI/ML are revolutionizing the field of networks. These technologies are being used to improve the performance, security, and reliability of networks.

Here are some of the key trends in AI/ML for networks:

  1. Simplify and scale data operations.

    AI/ML can be used to automate and simplify many of the tasks involved in managing and analyzing network data. This can free up network administrators to focus on more strategic tasks.

  2. Increase accuracy of forecasts.

    AI/ML can be used to predict network traffic patterns, identify potential problems, and plan for future capacity needs. This can help organizations to avoid costly downtime and improve the quality of service for their users.

  3. Decrease time to market.

    AI/ML can be used to automate the process of designing, deploying, and managing new network services. This can help organizations to bring new products and services to market faster.

  4. Enable insights on otherwise unusable data

    AI/ML can be used to extract insights from network data that would otherwise be too complex or voluminous to analyze manually. This can help organizations to identify security threats, optimize network performance, and improve customer experience.

    Figure-3: Trends in ML

AI/ML Use Cases

The introduction of AI/ML use cases in network functions has revolutionized the field of networking. AI/ML technologies are being leveraged to enhance network security, optimize network design and performance, anticipate traffic demand, and automate network tasks. This integration leads to improved efficiency, reliability, and overall network performance.

Examples of the popular use cases of AI/ML in large networks.

Figure-4: AI/ML Use Case: Hardware Failure Prediction

Figure-5: AI/ML Use Case: Network Demand Forecasting

ML vs Non-ML Networks

The comparison of ML-based and non-ML-based networks provides valuable insights into the advantages and limitations of each approach. By examining the key aspects such as scalability, flexibility, accuracy, and security, organizations can make informed decisions about the most suitable solution for their specific networking needs. This comparison can guide network engineers, architects, and decision-makers in selecting the optimal approach to meet their performance, efficiency, and security requirements.

A comparison between ML-based and non-ML-based solutions is provided in the followingtable:

Figure-6: Comparison of ML and non-ML solutions

Reasons Not to Use AI/ML

While AI/ML technologies offer significant benefits for networks, there are certain scenarios where their application may not be suitable or feasible. Several factors, such as data availability, use case definition, cost considerations, the need for customized models, and the effectiveness of existing automation, can influence the decision to refrain from using AI/ML in networks. Understanding the limitations and potential drawbacks of AI/ML is crucial for organizations to make informed choices about the most appropriate approach for their specific networking needs.

  1. Not enough data sets to train the model:
    • AI/ML models require large amounts of high-quality data to train effectively. In the context of networks, it may be challenging to collect and prepare sufficient data. Factors such as network size, traffic patterns, and security considerations can make data collection a complex and time-consuming process.
    • The lack of adequate data can lead to models that are not well-generalized and may not perform well in real-world scenarios.
  2. Use case is not defined well:
    • AI/ML models are designed to solve specific problems or achieve specific goals. If the use case for AI/ML in networks is not clearly defined, it can be difficult to develop a model that effectively addresses the desired outcomes.
    • A poorly defined use case can lead to misalignment between the model’s capabilities and the actual requirements of the network.
  3. High cost is a problem:
    • Implementing AI/ML solutions in networks can be expensive. Factors such as hardware requirements, software licenses, and the cost of hiring skilled professionals contribute to the overall cost.
    • Organizations need to carefully evaluate the cost-benefit analysis before investing in AI/ML for their networks. In some cases, the cost of deploying and maintaining an AI/ML solution may outweigh the potential benefits.
  4. Customized AI/ML model is required:
    • Off-the-shelf AI/ML solutions may not always be suitable for specific network scenarios. Organizations may require customized models that are tailored to their unique requirements.
    • Developing customized AI/ML models requires specialized expertise and resources, which can further increase the cost and complexity of the project.
  5. Existing automation is already serving the requirement:
    • Many networks already have existing automation solutions in place, such as network management systems (NMS) and configuration management tools. These solutions provide a range of automation capabilities that may already be sufficient for the organization’s needs.
    • Implementing AI/ML in such scenarios may not offer significant additional benefits or may require a substantial investment to achieve incremental improvements.

AI/ML Challenges in Networks

AI/ML in networks has benefits but also challenges. Complexity arises from numerous interconnected components and interactions, which AI/ML further complicates. Data limitations and algorithmic bias are additional concerns. Regulatory compliance adds another layer of complexity. Some of the challenges are described in detail below:


  • As networks become increasingly complex, it can be difficult to troubleshoot issues that arise. This is due to the large number of interconnected components and the complex interactions between them.
  • For example, a problem with a single router can have a cascading effect on the entire network, making it difficult to identify the root cause of the issue.
  • Additionally, the use of AI and ML in networks can further increase complexity by introducing new layers of abstraction and decision-making.

Data Requirements

  • AI and ML algorithms require large amounts of data to train and operate effectively. This can be a challenge for networks, as they may not have access to sufficient data to train their models.
  • For example, a network security system may not have enough data on recent attacks to train a model to detect and prevent future attacks.
  • Additionally, the data that is available may be biased or incomplete, which can lead to inaccurate or unfair models.

Algorithmic Bias

  • AI and ML algorithms can be biased, which can lead to unfair or discriminatory outcomes. This is because the algorithms are trained on data that may contain biases, such as racial or gender bias.
  • For example, a facial recognition system may be biased towards certain ethnicities, leading to false identifications or denials of service.
  • It is important to address algorithmic bias in networks to ensure that AI and ML are used in a fair and responsible manner.

Regulatory Compliances

  • Networks are subject to a variety of regulatory compliance requirements, such as the General Data Protection Regulation (GDPR) and the Health Insurance Portability and Accountability Act (HIPAA).
  • These regulations impose strict requirements on how data is collected, stored, and used.
  • AI and ML can add additional complexity to compliance, as they can introduce new data processing and decision-making processes.
  • Organizations need to carefully consider the regulatory implications of using AI and ML in networks to ensure that they are compliant with all applicable regulations.

Ethical Concerns

  • The use of AI and ML in networks raises several ethical concerns, such as the misuse of data and job replacement.
  • For example, AI-powered surveillance systems could be used to track and monitor people without their consent, raising concerns about privacy and civil liberties.
  • Additionally, AI and ML could lead to job automation, which could displace workers and have a negative impact on the economy.
  • It is important to consider the ethical implications of using AI and ML in networks to ensure that they are used in a responsible and ethical manner.

Networks: AI/ML Benefits

In today’s digital world, networks are becoming increasingly complex and interconnected. To manage and operate these networks effectively, organizations are turning to AI/ML. AI/ML can automate repetitive tasks, identify, and mitigate network threats, and optimize network performance. AI/ML can also help organizations to gain more insights from their network data, which can lead to better decision-making and improved business outcomes. Some of the top benefits are described below:

Lower Cost:

  • Automated tasks: AI/ML can automate repetitive and time-consuming network tasks, such as configuration, monitoring, and troubleshooting. This can free up staff to focus on more strategic initiatives.
  • Efficient customer support: AI/ML-powered chatbots and virtual assistants can provide 24/7 customer support, answering common questions and resolving simple issues. This can reduce the need for human customer support representatives, saving costs.
  • Improved performance: AI/ML can be used to optimize network performance by identifying and resolving bottlenecks and inefficiencies. This can lead to reduced latency, improved throughput, and better overall network performance while minimizing the network operation cost.

Reduced Network Risk:

  • Resilient network: AI/ML can be used to create more resilient networks that are better able to withstand outages and attacks. This can be done by predicting and preventing network failures, and by quickly identifying and resolving issues.
  • Identify and mitigate threats: AI/ML can be used to detect and mitigate network threats, such as malware, DDoS attacks, and phishing attempts. This can help to protect sensitive data and systems from being compromised.
  • Accurate network trends and forecast: AI/ML can be used to analyze network data to identify trends and forecast future needs. This information can be used to make informed decisions about network planning and investment.
  • Network outage prediction: AI/ML can be used to predict network outages before they occur. This can help to prevent downtime and lost productivity.

More Revenue:

  • Enhanced network and capacity planning: AI/ML can be used to optimize network and capacity planning, ensuring that the network has the resources it needs to meet current and future demands. This can help to avoid costly over-provisioning or under-provisioning of network resources.
  • Faster time to market: AI/ML can help to accelerate time to market for new network services and applications. This can be done by automating the testing and deployment process, and by identifying and resolving potential issues early on.
  • Better customer experience: AI/ML can be used to improve the customer experience by providing personalized and proactive support. This can lead to increased customer satisfaction and loyalty.

Networks: AI/ML Innovation Catalysts

The convergence of AI/ML with networks is revolutionizing various industries. Here are some key factors driving this transformation:

  1. Increase in Data/Compute and Storage:
    • The proliferation of IoT devices has led to an exponential growth in data generation, fueling AI/ML innovation.
    • High-performance computing (HPC) clusters and cloud platforms provide the necessary compute and storage resources for complex AI/ML models.
  2. Edge Computing:
    • Edge computing brings AI/ML capabilities closer to data sources, enabling real-time decision-making.
    • Edge devices, such as sensors and gateways, collect and process data locally, reducing latency and bandwidth requirements.
  3. Cloud Infrastructure:
    • Cloud platforms offer scalable and elastic infrastructure for deploying and managing AI/ML workloads.
    • Cloud-based AI/ML services provide pre-built tools and frameworks for developers, accelerating the development and deployment of AI/ML applications.
  4. Increase in Devices Running AI:
    • Smartphones, smart home devices, and autonomous vehicles are increasingly equipped with AI capabilities.
    • These devices generate vast amounts of data and use AI to perform tasks such as image recognition, natural language processing, and predictive analytics.
  5. Pre-trained Models:
    • Pre-trained models, such as open-source BERT and ResNet, provide a starting point for developing custom AI models.
    • These models have been trained on large datasets and can be fine-tuned for specific tasks, reducing the time and resources required for model development.
  6. Human and AI Cooperation:
    • AI/ML is augmenting human capabilities, enabling collaboration between humans and machines.
    • Human-AI teams can leverage their respective strengths to solve complex problems and make better decisions.


AI and ML are revolutionizing the field of networking, bringing efficiency, automation, and significant performance improvements. As networks continue to grow and complexity, traditional management methods are becoming increasingly ineffective. AI and ML offer a powerful solution by enabling networks to self-configure, self-optimize, and self-heal, leading to a more agile, resilient, and cost-effective network infrastructure. The use of AI and ML in networks is still in its early stages, but it has the potential to transform the way networks are designed, built, and operated. As AI and ML technologies continue to evolve, we can expect to see even more innovative applications that will further unleash the potential of networks.



**** This blog post was written with the assistance of Google’s Gemini. The AI was used to generate initial draft, rephrasing, and brainstorming, which I then refined, edited, and expanded upon.

Part1: Unleashing Network Potentials: Current State and Future Possibilities with AI/ML

By Vinay Tripathi with Ajay Lotan Thakur


We live in an era of rapid digitization and ubiquitous connectivity where networks touch every aspect of our lives. From the global telecommunication infrastructure enabling seamless voice and data communication to the diverse social media platforms facilitating instant global interactions, the way we collaborate, communicate, and access information is heavily dependent on the seamless operation of networks. However, as networks continue to evolve, expanding in size and complexity, managing, provisioning, and optimizing them efficiently poses significant challenges.

Introducing Artificial Intelligence (AI) and Machine Learning (ML), offering a transformative solution to simplify network provisioning, streamline operations, enhance network performance, and unlock valuable insights from the vast amounts of network data. AI and ML empower network administrators, architects, planners, engineers, and managers with a range of capabilities that significantly improve network efficiency and effectiveness.

Type of Networks

Networks can be classified into various types based on their purpose, size, and geographical coverage. Some common types of networks include:

1. Core Networks:

  • Form the backbone of the internet, connecting large geographical regions and major network providers.
  • Characterized by high-speed data transmission, typically fiber optic cables, and redundant paths for reliability.
  • Responsible for routing traffic between different parts of the internet and carrying large amounts of data.

2. Data Center Networks:

  • Designed to support the infrastructure of data centers, where large amounts of data are processed and stored.
  • Highly interconnected and optimized for low latency and high bandwidth to facilitate efficient communication between servers and storage systems.
  • Often utilize specialized networking technologies such as Ethernet and InfiniBand.

3. Enterprise Networks:

  • Connect devices and resources within an organization or company.
  • Include local area networks (LANs) for devices within a building or campus and wide area networks (WANs) for connecting geographically dispersed sites.
  • Provide secure and reliable connectivity for employees, customers, and partners.

4. Cellular Networks:

  • Provide wireless connectivity for mobile devices such as smartphones, tablets, and IoT devices.
  • Consists of cellular towers or base stations that communicate with mobile devices using radio waves.
  • Offer various cellular technologies such as 2G, 3G, 4G, and 5G, each providing different levels of speed and capacity.

Here’s an example that demonstrates various types of networks:

Figure-1: Various types of Networks

Figure-2: Google Network Infrastructure

Network Functions

Networks are designed to achieve specific goals, for example, edge networks can have very different routing and switching requirements when compared to core networks. However, there are some functions which are common to all networks.

  • Engineering:Deals with design, optimization, provision and development of the network infrastructure and services. Engineering teams ensure the network operates efficiently, reliability and meets all the performance and scale requirements.
  • Capacity Planning & Forecasting:Estimates future demand of network resources such as routers, switches, servers, storage and bandwidth. It helps in network planning and scaling by analyzing history consumptions and future demand.
  • Implementation:Physically deploys the network components like routers, switches, servers, etc. It integrates the systems to the rest of the network and services based on the designs and plans developed by the engineering team.
  • Monitoring:Another critical function of network infrastructure which provides vital insights to the current state of network infrastructure. Data collected from the systems can be used by other network functions to improve network performance, reliability, and security.
  • Operation:A crucial function of the network which focuses on day-to-day management, maintenance and support of network infrastructure and services. It ensures the network operates smoothly, efficiently and with least disruptions.
  • Security:Maintains confidentiality and integrity of information and systems. It uses firewall, intrusion detection systems and access control lists to keep the network secure.

Network Without AI/ML

Many large-scale network outages result from human manual errors or automated system malfunctions. Avoiding such issues is difficult when humans are involved in daily operational decision-making. Many network functions have been automated in recent years, but they still rely on predefined values or actions that require continuous system or service updates. Additionally, there are still many networks or functions that are not automated due to a lack of expertise, resources, or willingness. Even in automated networks, operators must perform manual operations in certain situations, such as tooling infrastructure failures or recoveries. Some scenarios where automated and/or manual operations are performed in a network include:

  • Manual/automated security provisions:
    • Manual security provisions involve tasks such as manually configuring firewalls, intrusion detection systems, and other security devices.
    • Automated security provisions involve using software tools to automate security tasks, such as vulnerability scanning, patch management, and threat detection.
  • Manual/automated configuration of network devices (switches, routers, etc.):
    • Manual configuration involves manually configuring network devices, such as switches and routers, using command-line interfaces or web-based interfaces.
    • Automated configuration involves using software tools to automate the configuration of network devices, which can save time and reduce errors.
  • Manual/automated monitoring dashboard with predefined values:
    • Manual monitoring involves manually monitoring network performance and security metrics using technologies such as Telemetry, SNMP, and syslog.
    • Automated monitoring involves using software tools to automate the monitoring of network metrics and generate alerts when predefined thresholds are exceeded.
  • Manual/automated troubleshooting of network issues:
    • Manual troubleshooting involves manually diagnosing and resolving network issues, such as connectivity problems, performance issues, and security breaches.
    • Automated troubleshooting involves using software tools to automate the diagnosis and resolution of network issues, which can reduce the time it takes to resolve problems.
  • Manual/automated mitigation of network events:
    • Manual mitigation involves manually responding to network events, such as security breaches, denial-of-service attacks, and natural disasters.
    • Automated mitigation involves using software tools to automate the response to network events, which can help to minimize the impact of these events.
  • Manual/automated capacity planning process:
    • Manual capacity planning involves manually forecasting network traffic demand and planning for future capacity needs.
    • Automated capacity planning involves using software tools to automate the forecasting of network traffic demand and the planning of future capacity needs, which can help to ensure that the network has sufficient capacity to meet future demand. Automated solutions can save time, reduce errors, and improve efficiency.

NextGen Network Requirements

Next-generation networks must meet diverse use cases and deliver exceptional customer experiences. Network applications and use cases constantly evolve, necessitating adjustments in network design, technologies, and operations. Continuous optimization is needed to unleash the network’s full potential. For example, existing data center networks require redesign and optimization to meet the demands of AI/ML applications. Critical requirements that must be fulfilled by next-generation networks are as follows:

  1. Increased performance, reliability, and security: Networks must handle massive data volumes and complex workloads with high performance and low latency. Reliability and security are paramount, ensuring uninterrupted operations and safeguarding sensitive information.
  2. Customer-centric focus: Delivering a seamless and delightful customer experience is crucial. Networks must facilitate seamless coordination across business functions, enabling personalized services and addressing customer needs effectively.
  3. Managing massive complexity: The convergence of 5G, Internet of Things (IoT), AI/ML loads and edge computing introduces unprecedented complexity. Networks need to be equipped with advanced orchestration and management capabilities to handle this complexity efficiently.
  4. Value beyond connectivity: Networks should not be limited to providing mere connectivity. They must deliver value-added services and capabilities such as real-time analytics, edge computing, and network slicing to meet diverse customer requirements.
  5. Improved service assurance and issue prediction: Networks must proactively monitor and analyze network performance to predict potential issues before they impact customers. Fault detection and self-healing mechanisms are essential to ensure uninterrupted service availability.
  6. Measuring and optimizing customer experience: Networks should have built-in capabilities to measure and analyze customer experience metrics such as latency, packet loss, and jitter. This data can be leveraged to optimize network performance and rectify areas of improvement.
  7. Understanding customer expectations: Networks must provide insights into customer expectations and evolving needs. This can be achieved through surveys, feedback mechanisms, and real-time monitoring of customer interactions.
  8. Increased efficiency and intelligence: Networks should incorporate AI and ML technologies to automate tasks, optimize resource allocation, and enhance overall network efficiency and intelligence.


Future networks need AI/ML integration to fulfill the next generation of requirements. AI/ML can make networks more efficient, secure, reliable, and scalable. AI/ML can effectively monitor and alert operators, utilizes resources efficiently, make network customer centric and faster delivery of services. In the next blog, we will discuss AI/ML use cases, benefits, limitations, and projections.



**** This blog post was written with the assistance of Google’s Gemini. The AI was used to generate initial draft, rephrasing, and brainstorming, which I then refined, edited, and expanded upon.

Building and Operating a Cloud Native 5G SA Core Network

By Ajay Thakur with Alan J Weissberger


In this article, we endeavor to clarify some of the critical issues and questions related to implementing a cloud native 5G SA core network and how it differs from the traditional core network composed of hardware devices and software solutions. It’s important to note that NONE of the 3GPP defined 5G features can be realized without a 5G SA core. Those include: Network Automation, Network Function Virtualization, 5G Security, Network Slicing, Multi-Access Edge Computing (MEC), Policy Control, Network Data Analytics, etc.

Communication Service Providers (CSPs) will need to do things differently (than 4G) in order to implement and use a 5G cloud native SA core. Various cloud native 5G SA core aspects include network planning, deployment, software upgrades, network monitoring, hardware and platform upgrades.

These will be examined and contrasted with traditional implementations, such as the 4G Evolved Packet Core (EPC).


3GPP introduced 5G SA core network architecture in release 15. Since then numerous new features (work items) have been introduced to specifications. 3GPP’s 5G SA’s specifications use virtualization and cloud native principles as the foundation. A few key 3GPP Technical Specifications (TSs) for 5G system are the following:

  • TS 22.261, “Service requirements for the 5G system”.
  • TS 23.501, “System architecture for the 5G System (5GS)”
  • TS 23.502 “Procedures for the 5G System (5GS)
  • TS 32.240 “Charging management; Charging architecture and principles”.
  • TS 24.501 “Non-Access-Stratum (NAS) protocol for 5G System (5GS); Stage 3”
  • TS 38.300 “NR; NR and NG-RAN Overall description; Stage-2”
  • TS 23.527 “5G System; Restoration procedures Stage-2”

5G SA tries to resolve the challenges faced by network operators in the EPC deployments and how those challenges can be mitigated with new design.

Several important changes in the 5G SA core are support for a Service Based Architecture and Cloud Native implementation of 5G SA core. That will enable new 5G features and functions like Network Slicing, 5G Security, and MEC, among others.

To reap the benefits of Cloud Native SA, CSPs are required to adapt to new cloud native principles of network deployment, operation and monitoring. We shall examine various aspects of Life Cycle Management of 5G SA software and also ask some open ended questions on each aspect.

SBA architecture Diagram with multiple Interfaces:

  • The Network diagram above shows the SBI interfaces in case of roaming. Each of the individual NF may be composed of one more micro-services. Each NF may come from different vendor. Since these NFs are available in containerized format, they may or may not share the same container orchestration platform.

Network Dimensioning:

  • 5G SA solves the network expansion problem since the network can be scaled up/down by adding/removing commercial off the shelf hardware.
  • Once we have all the NFs software releases available, operators can gather the compute, memory & network requirements for deployment.
  • Operators would rely on auto scaling functionality provided by the 5G NF vendors to avoid over provisioning at the start. This is relatively much simpler compared to adding dedicated hardware for NF in case of EPC.

Deployment Options Selection/Planning

  • Operators need to decide the type of deployment for the network like the 5G SA deployment on public cloud or on private cloud.
  • Along with Public vs Private cloud decisions, the Operator also needs to decide on a Container orchestration engine. Container orchestration can be managed service or operator managed service. Popular container orchestration engine is Kubernetes (K8s).
  • Next step would be to finalize or select one of the CI/CD tools which works for all the vendors and integrate that with the container orchestration platform.
  • Also placement of the NFs needs to be decided, e.g. User Plane Function (UPF) could be on-prem close to RAN. It’s possible that the operator may place some of the RAN virtualized components in the cloud.
  • These all are important decisions to be made before getting into the real deployment of the software.
  • Some of the aspects here are due to Cloud Native 5G SA and were not applicable for EPC.


  • It can be seen from the CNCF project that there are many projects which are available for CI/CD
  • Irrespective of Public or Private cloud, operators need to follow cloud native CI/CD principles for deploying the Core Network in the Cloud environment. CI/CD involves taking the software release from the vendor and running it against existing network functions and carrying out some minimum tests and once operator is satisfied with evaluation of the release then rollout the release in the network.
  • Operators can decide to have a separate staging environments where new releases can soak for a certain period of time while few subscribers use the new release.
  • CI/CD gives the option of rolling back the release if the operator is not happy with the performance of the new release.
  • Now the challenge here would be, will the operator have single CI/CD tools used for deploying all the vendors’ solution OR would the operator use its own CI/CD tooling and integrate NFs from vendor it in its own environment. This is the decision the operator needs to take.
  • Having Automated CI/CD infrastructure which takes the software releases from the vendors and passes it through multiple environments and all the way to the production environment is key.
  • Without having an appropriate CI/CD environment it would be very difficult to manage all micro-services and their deployments.
  • Public Cloud may come with inbuilt CI/CD solutions and would be easy for operators to start with.
  • AWS offers multiple services around CI/CD and those are listed here –
  • Azure offering can be found here
  • Google Cloud CI/CD services can be found here –

Software Upgrade:

  • Cloud Native 5G SA allows operator to upgrade some of the components easily instead of complete 5G Core update in one go. CI/CD framework would help in the software upgrades with minimal human intervention.
  • Note that with Cloud Native principles the operator would get multiple patch/minor releases and may be some major releases throughout the life cycle of the software. So the traditional approach of pulling down complete hardware & upgrading it with new software is not required for microservice based solutions. But this really works as long as all microservices are truly built stateless and supports live upgrade.
  • As an operator, it would be required to know the impact of each upgrade package and prepare for rollback in case something goes wrong during the upgrade cycle.

Network Monitoring:

  • Traditionally, operators developed their own Network Monitoring solutions to monitor the health of EPC since there was no standard mechanism to get the metrics, statistics from the NFs,
  • 5G SA follows Cloud Native principles; it is easy to get the logs, statistics, alerts, alarms from all microservices in a consistent manner. There are common tooling used by most of the cloud native applications and CNCF has multiple projects in these categories.
  • CNCF supported Monitoring projects are shown below figure

    • Tracing is important aspect to find out the bottleneck in the performance.


  • Logging has been a traditional approach for debugging network issues. Below are the projects offered by CNCF in the logging area

  • Public Cloud providers can extend the monitoring easily by generating Texts or email alerts as per CSPs needs.
  • Operators can define their policies to retain the network performance monitoring output for a long time and can take backup of this easily through use of Public Cloud Providers data backup services.
  • In the case of EPC these mechanisms were product dependent.
  • Challenge in case of 5G SA would be each NF vendor may end up in using different tool and operators would have some challenges to converge all NF vendors to the common tooling.

Hardware & platform upgrade:

  • In the traditional approach, providing hardware with updated operating systems and platforms was the responsibility of the equipment vendor. Now this responsibility has gone into either operator’s hand or sometimes in Public Cloud Provider’s hand. It depends on if the 5G SA is deployed on on-prem or on public cloud. Operators need to carefully plan for these upgrades without causing any downtime and of course follow the rolling upgrade patterns to avoid updating multiple entities at a time.
  • If managed container orchestration is used, then these upgrades are seamlessly handled by Public Cloud Providers.

Vendor Lock In:

  • In the case of EPC, vendor lock in was specific to NF & Radio Vendors. The 3GPP EPC specification allows the operators to swap NFs from one vendor with another and as long as NF supports the required Services. Point to note that this is less costly replacement compared to replacing one vendor with another vendor when NF had associated hardware.
  • But with cloud native SA there is a chance that the operator may end up in building the tooling (CI/CD, monitoring etc.) over the period of time and this may lead to cloud provider lock in.


The 5G SA core network provides a lot of flexibility and automation via cloud native deployments. However, the 3GPP 5G core specs contain a lot of implementation options, which network operators and their vendors must select to properly deploy a 5G core network. Making those decisions will likely require solid experience with operating applications on a cloud native platform. And that may be a reason that 5G SA core network rollouts have been so slow.

In the U.S. AT&T and Verizon have taken a very cautious approach to deploying their long ago promised 5G SA core networks. During the Brooklyn 6G Summit in November 2023, Chris Sambar, EVP for technology at AT&T said, ““I would say we are not moving as quickly as some of the other operators on the 5G standalone core, but we see the use cases that are coming, we understand when they’re coming, so we’re being very purposeful about getting there when we need to get there.” That’s despite AT&T outsourcing its 5G SA core network deployment to Microsoft Azure in June 2021. Yet Microsoft is the world’s second biggest cloud services provider with tons of experience running cloud native applications.

Summing up, Dave Bolan, Research Director at Dell’Oro Group wrote, “The buildout of 5G SA networks is going slower than anticipated which is restraining growth in the marketplace. To date, we count fifty 5G SA eMBB (enhanced Mobile BroadBand) networks that have been commercially deployed worldwide by Mobile Network Operators (MNOs). We counted 18 new 5G SA networks in 2022, but only 12 were launched in 2023. On a positive note, we believe a lot of work has been done in the background, preparing for 5G SA launches by Mobile Network Operators (MNOs) and we expect 2024 to have more launches than 2022.”


Ajay Lotan Thakur,  Cloud Software Architect at Intel and IEEE Techblog Editorial Team member








Global 5G Market Snapshot; Dell’Oro and GSA Updates on 5G SA networks and devices

Ericsson Mobility Report touts “5G SA opportunities”

Analysys Mason: 40 operational 5G SA networks worldwide; Sub-Sahara Africa dominates new launches

Samsung and VMware Collaborate to Advance 5G SA Core & Telco Cloud

5G SA networks (real 5G) remain conspicuous by their absence

GSM 5G-Market Snapshot Highlights – July 2023 (includes 5G SA status)



Top Optical Network Equipment Vendors: Data Center Interconnect & Overall Market

Executive Summary:

Market research firms Dell’Oro and Heavy Reading disagree on who are the top optical network equipment vendors, especially for data center interconnect (DCI).  Obviously, the mega cloud computing/Internet service providers (Google, Amazon, Baidu, Facebook, etc) together account for the overwhelming market for DCI equipment purchased.  None of them disclose who their DWDM vendors are.  It’s well known that most of those mega cloud/Internet players design their own IT equipment (e.g. compute servers, Ethernet switches, Routers, etc), but they don’t design or build DWDM transport gear.

Dell’Oro Group DCI Market Analysis:

Ciena, Cisco and Infinera together command 85% of the disaggregated wavelength-division multiplexing field for DCI optical network equipment market segment, Dell’Oro Group estimates.


–>This is a big surprise to this author as neither Nokia (via Alcatel-Lucent), Huawei (#1 overall optical network vendor) or ZTE are top tier according to Dell’Oro.  See two graphs below (“Other Voices” section), courtesy of Heavy Reading and IHS-Markit.


Dell’Oro Group estimated that disaggregated WDM systems reached an annualized revenue run rate of $400 million, growing 225% year-over-year. This is partly because these systems are finding utility in the booming DCI market segment.

Jimmy Yu, VP at Dell’Oro Group, said that while the disaggregation concept is not new, service provider adoption in the data center segment is.

“In most—if not all—purchases, we found that these new systems were being employed in DCI across both metro and long haul spans,” Yu said in a press release. “So far, the largest consumers have been internet content providers that appreciate the platform for its simplicity, capacity, and power savings.”

Yu added that “based on second quarter results, where disaggregated WDM systems represented nearly one-third of the optical DCI equipment purchases made, we have to say that Disaggregated WDM systems are truly hitting the sweet spot for DCI.”

As wireline operators look to diversify their revenue mix, the DCI market has a compelling growth path driven by the consumption and distribution of various data forms over the public internet and private networks.

Outside of DCI, the overall WDM market, which consists of WDM Metro and DWDM Long Haul, grew only 2% year-over-year in the second quarter, says Dell’Oro. The research firm noted that growth was driven by strength in the Asia Pacific region, especially China and India.

The share of 100G WDM wavelength shipments going to DCI was 14% in the quarter, according to Dell’Oro.

About the Report:

The Dell’Oro Group Optical Transport Quarterly Report offers complete, in-depth coverage of the market with tables covering manufacturers’ revenue, average selling prices, unit shipments (by speed including 40 Gbps, 100 Gbps, and >100 Gbps).  The report tracks DWDM long haul terrestrial, WDM metro, multiservice multiplexers (SONET/SDH), optical switch, optical packet platforms, and data center interconnect (metro and long haul).  To purchase this report, call Matt Dear at +1.650.622.9400 x223 or email [email protected].


Other Voices on Optical Network Equipment Market:

1.  Cignal AI:

Huawei and ZTE saw record shipments of 100-Gbps coherent ports in China during the second quarter of 2017 as well as strong sales in general throughout the region, reports Cignal AI. So what accounts for sour grapes from optical component houses? Inventory corrections at Chinese systems vendors, particularly Huawei, according to the market research firm.

“Demand for optical hardware in China is not slowing down, and equipment vendors are universally providing positive guidance for North America during the second half of the year,” said Andrew Schmitt, lead analyst for Cignal AI. “Operators around the world are shifting spending from long-haul to metro WDM, though this shift is materializing into gains for only a few vendors.”

Optical revenue in China is up 13 percent for the first half of 2017 as compared to the same period in 2016. The weak demand reported by component makers is a result of an ongoing inventory correction (primarily at Huawei), rather than a signal of weak end market demand.

2. IHS-Markit:

Huawei ranked first overall in combined market presence and market leadership in the recent Optical Network Hardware Vendor Scorecard released by IHS Markit.  Huawei received this assessment for its comprehensive performance on multiple benchmarks including reputation for innovation, market share momentum, and global market share.

There are over a dozen vendors around the globe that make and sell optical network equipment. The 10 vendors profiled in this Scorecard–ADVA, Ciena, Cisco, Coriant, ECI, Fujitsu, Huawei, Infinera, Nokia, and ZTE–were selected because they are the top revenue producers of optical hardware.

The Scorecard used concrete data and metrics, including market share, financials and direct feedback from buyers on innovation, product reliability, service and support to evaluate 2016 market performance and future momentum of the top 10 optical network equipment vendors.

IHS Markit optical network hardware vendor scorecard (Source: IHS Markit Optical Network Hardware Vendor Scorecard)

3.  Heavy Reading:

Market share estimates are based on DCI revenue contribution by Heavy Reading’s definition (not disclosed in the teaser briefing). Most vendors do not currently break out from their broader metro WDM revenue the portion accounted for by metro DCI deployments. A few companies did provide Heavy Reading with some general guidance on their revenue from metro DCI. The pie chart figure below shows Heavy Reading’s metro DCI equipment vendor share estimates for 2016.

Note that Adva has the top vendor market share and Cisco is not represented in the figure.


Source:  Heavy Reading




AT&T Moves Aggressively on & Expansion of its Fiber Network

AT&T Expands & FTTH Deployments:

In sharp contrast to Verizon’s decision NOT to deploy, AT&T has announced expansion of its service for multi-dwelling units (MDUs) and its fiber-to-the-home network (AT&T Fiber).

The mega telco will extend its all-fiber network in two markets — Biloxi-Gulfport, MS and Savannah, GA.  AT&T will also be offering its hybrid fiber-coax service for MDUs in 22 metropolitan markets.

The AT&T deployments will use “fiber runs to the telecom closet on the property, and individual coax runs to each apartment unit,” an anonymous AT&T spokesperson said to Telecompetitor.

Residents of properties served will also be able to obtain DIRECTV service without installing a dish at their individual units. Instead, the video service will be delivered over D2 Advantage, which the AT&T spokesperson described as “a centrally wired satellite dish that is shared among residents in the property.”

AT&T announced eight metro areas where can be deployed immediately, including Boston, Denver, Minneapolis, New York City, Philadelphia, Phoenix, Seattle and Tampa. In 14 other markets, consumers in target MDUs can order service now for deployment in “the near future,” the company said.

AT&T is one of multiple carriers that are looking at as part of their broadband strategy. The technology can support considerably higher speeds than DSL or fiber-to-the-neighborhood (FTTN) services – and although bandwidth is lower than it might be for a fiber-to-the-home deployment, the cost is considerably less.

The news that AT&T is deploying is not surprising, as the company already has conducted a trial of the service in Minneapolis and executives have indicated deployment plans.  At this year’s Open Network Summit (ONS), AT&T’s Tom Anschutz told an audience that would improve the speed and signal quality of data transmission on older, low grade twisted pair, which is used in many MDUs and in condominium complexes (where this author lives).  He hinted that market segment would be a focus area for AT&T.


AT&T is extending the reach of its fiber network:

AT&T claims to have the largest fiber network in its 21-state home broadband footprint, reaching more than 5.5 million residential and commercial locations across the 57 markets after adding over 1.5 million sites since January 1st. Plans call for extending service availability to another 1.5 million locations by year’s end, boosting the total to 7 million.

Of those 5.5 million homes and businesses now reached by AT&T Fiber, the mega telco said it has signed up more than 2 million broadband subscribers. The company did not, however, break out how many of those subs are new ones, as opposed to DSL customers who have been upgraded to the new FTTH network.

AT&T  is the US’s third-largest broadband provider after Comcast Corp and Charter Communications Inc with nearly 15.7 million subscribers at the end of June, 2017.

However, the mega telco ranks #1 on Vertical Systems U.S. Fiber Lit Buildings (Fiber to commercial buildings) leaderboard:


Huawei & Smart Axiata launch 4.5G (???) network in Cambodia

Executive Summary:

Cambodia’s mobile telecommunications company Smart Axiata, in partnership with global ICT solutions provider Huawei, launched the 4.5G [1] network in Cambodia on Monday, August 21st.

Cambodian Minister of Posts and Telecommunications Tram Iv Tek said Smart Axiata was the first telecom operator that introduced 4.5G evolution technology to Cambodia.  Tram said:

“This latest evolution of mobile technology will enable subscribers to enjoy better and faster mobile internet,” he said during the launching ceremony. “Fast access to the internet is indeed an important tool in developing a digital economy.”

Thomas Hundt, chief executive officer of Smart Axiata, said 4.5G network is capable of providing an Internet speed 8 times higher than that of 4G.  [However, he didn’t state what attributes 4.5G has and what, if any standard it’s based on.]

“We strive to be the most progressive telco in the country. Smart Axiata will continue to seek new and better ways to ensure our subscribers receive the best possible service,” he said. “After being the first telco introducing 4G LTE in Cambodia in 2014, 4G+ in 2016, and now 4.5G, and 5G is surely not far away.”


[1]  What is 4.5G?

There is no official definition of 4.5G by any standards body or forum.  ITU-R doesn’t even have a definition of 4G.  We assume that 4.5G is anything better than LTE-Advanced (as defined by 3GPP).  Note that ITU-R had originally said that LTE Advanced was 4G, before marketing organizations hijacked the term to mean anything better than the 3G network they had deployed.

Huawei’s website says there are three core concepts in 4.5G: Gbps, Connection+, and Experience 4.0. The company says that “4.5G further increases data rates for better user experience and expands applications in vertical industries. This helps operators create new business applications in vertical industries. This helps operators create new business opportunities and gain a competitive edge in the next few years.”

Huawei notes that “3GPP approves LTE-Advanced Pro as New Marker for LTE Evolution System (4.5G).”

Meanwhile, the 3GPP website doesn’t refer to LTE Advanced Pro as 4.5G.  It states:

LTE-Advanced Pro will allow mobile standards users to associate various new features – from the Release’s freeze in March 2016 – with a distinctive marker that evolves the LTE and LTE-Advanced technology series.

The new term is intended to mark the point in time where the LTE platform has been dramatically enhanced to address new markets as well as adding functionality to improve efficiency.


Promise and Potential of 4.5G:

Li Xiongwei, chief executive officer of Huawei Technologies Cambodia, said Huawei was very pleased to cooperate with Smart to launch 4.5G and to bring Cambodia closer to next generation technology.

“The 4.5G will certainly bring significant network improvements from today’s 4G network, while acting as a bridge for future mobile data applications when 5G is launched in the future,” he said.

Cambodia has six mobile phone operators and about 30 Internet service providers.

According to Cambodia’s Ministry of Posts and Telecommunications, approximately 8 million out of the country’s 15 million people have access to the Internet.  About 45 percent of the population is still offline, the minister said, adding that the country expected that 100 percent of the urban dwellers and 80 percent of the rural dwellers would use the Internet by 2020.

Huawei Partners with Smart Axiata to Launch the First 4.5G Network in Cambodia


Huawei’s 5G oriented X-Haul:

On August 14th, Huawei officially released its 5G-oriented mobile bearer solution called X-Haul.    The new fronthaul and backhaul network has four core values: providing flexible access capabilities that can match the scenario of any site; implementing agile network operations based on a cloud architecture; enabling new service innovation through end-to-end network slicing; and supporting smooth evolution from 4G bearer networks to 5G bearer networks.

Flexible networking is implemented through IP, microwave, and OTN access technologies, enabling unified fronthaul and backhaul whether or not fiber optic cables are used.


1.  Via email, Thomas Hundt wrote:

“4.5G aka LTE Advanced Pro is the latest evolution of 4G networks, in line with 3GPP release 13 and 14. By using technologies such as 4×4 MIMO, 256QAM modulation the network capacity is significantly increased which leads to much faster end user speeds provided a 4.5G enabled phone is used.”

2.  Derek Kerton of the Telecom Council/Kerton Group wrote via email that Nokia has introduced 4.9G technology: Nokia Introduces 4.9G as Next Incremental Step to 5G




KT opens 100,000 free Wi-Fi Access Points in South Korea; WiBro & LTE

KT, the second largest telco in South Korea,  has opened up 100,000 Wi-Fi access points (APs) as part of its participation in a government-led program which will improve the public’s access to free Wi-Fi-based connectivity.  According to the Korea Times the operator has also pledged to enhance Wi-Fi network equipment in subway trains across the country.

“Aiming at reducing people’s telecom (i.e. broadband Internet access) expenses, we have worked on opening 100,000 Wi-Fi APs and improving the quality of Wi-Fi networks inside subway trains,” said Park Hyun-jin, vice president of KT Mobile’s business department. “We will come up with more measures to further cut household telecom expenses and expand benefits for our subscribers.”

With the government’s ‘Public Wi-Fi 2.0’ policy seeking to improve Wi-Fi APs nationwide, KT has said that the bulk of its newly opened hotspots are at busy locations, such as shopping malls, bus stations, subway stations and tourist sites. The hotspots can be accessed by both KT mobile subscribers and those that are not signed up to the cellco’s services, though the latter users are required to provide personal information (such as an email address, phone number, gender and age) in order to take advantage of free Wi-Fi for one hour.

Alongside its expansion of free Wi-Fi APs, KT has begun work on improving network equipment to provide faster and more stable wireless broadband connectivity. To that end, it is reportedly replacing old Wi-Fi network devices, which are based only on WiBro technology (the Korean variant of WiMAX), on subway trains with new hybrid devices that support both WiBro and LTE technologies. KT has said it expects to first complete the replacement for subway trains on lines 1 to 8 in Seoul, before expanding upgrade works nationwide by the end of this month.

KT models show promotion materials for the Giga-LTE WiFi network on June 15.

KT models show promotion materials for the Giga-WiFi and LTE networks ………………………………………………………………………………………………..

Over one year ago, KT introduced its GiGA WiFi 2.0 that will provide network speeds twice as fast as the previous version. With the world’s first Wi-Fi 2.0 technology, the nation’s second-largest telecom company said it will soon be able to upgrade maximum Wi-Fi speed to 3.4 gigabits per second (Gbps).

The Wi-Fi 2.0 network adopted “multi-user, multiple input and multiple output” technology to provide faster speeds to all users simultaneously connected to a network, KT said. It has also applied the “wireless intrusion prevention system” to block unauthorized access points and devices.
KT also said it has provided more than 200,000 public Wi-Fi access points nationwide of which 100,000 were announced this past Friday.


AT&T praises Intel’s role in network virtualization & 5G readiness

AT&T describes its relationship with Intel as “a push-pull strategy” as the huge U.S. telco invests in virtualization and upgrading to a 5G network with the aid of Intel’s Xeon Scalable processors.

“Without some of those advantages [from the new Xeon Scalable processors] and capabilities that have been created in the software space, we wouldn’t be able to do it,” said AT&T’s Chris W. Rice, SVP of AT&T Labs and Domain 2.0 architecture. “It is a key underpinning in our SDN-network virtualization journey. Intel pushed the technology into the ecosystem, the capabilities and the chips, and then we can pull it through the ecosystem.” Rice added.


Author Notes:

1.  AT&T buys Compute Servers which contain Intel Xeon processors:

It’s important to recognize that AT&T does NOT buy processor chips from Intel or any other semiconductor company.  It buys compute servers which contain Intel Xeon processors.  While the compute server vendor(s) have not been disclosed, it’s likely one or more Chinese or Taiwanese ODMs.

According to IDC, X86 machines dominated the compute server market in 2016.  Servers using mostly Xeon processors accounted for $11.2 billion in sales, down 3.1 percent. Server machines using other processor architectures, including Itanium, Power, Sparc, ARM, and a smattering of others, drove $1.3 billion in revenues, but fell 30 percent year on year.  Intel X86 compute hardware had a 99.2 percent shipment share and an 89.6 percent revenue share, said IDC in a research report.

An article summarizing IDC and Gartner Group reports on 1Q2017 compute server shipments is here.  The rise of ODMs is described in this blog post.

In January 2016, AT&T joined the Open Compute Project which is specifying open source hardware (e.g. compute servers and Ethernet switches) for use in data centers. AT&T has repeatedly stated it wants to make its Central Offices look like cloud resident data centers.

2.  AT&T’s Cloud & Virtualization Platforms:

The AT&T Integrated Cloud (AIC) is a data center design that includes top-of-rack switches, storage, servers, and software at the hypervisor. When complete, AIC will encompass more than 1,000 zones distributed around the globe.  AIC is based on the open source  OpenStack cloud management framework.

AT&T’s Universal CPE (uCPE) is the hardware foundation of its Network Functions on Demand service. It’s an AT&T-branded Intel x86 server (presumably made by a Chinese ODM) that sits at the enterprise premises and can mix and match software-based VNFs, depending on what functions are needed at each location. The uCPE was designed and manufactured to AT&T’s specifications to enable customers to run multiple VNFs on one device.


According to SDx Central, AT&T has deployed two workloads on Intel’s Xeon Scalable processors and says others are in the queue. The two workloads are AT&T’s virtual Content Distribution Network (vCDN) and its virtual VPN Internet Gateway (vVIG).

vVIG is a virtual machine that acts as an IPSec gateway between unsecure and secure networks, providing data security at the IP packet level. It uses Intel’s Data Plane Development Kit (DPDK) to speed up the cryptographic processing of IPSec data packets.

Using the new Intel processors allows the vVIG to support a larger data throughput for less cost and a smaller footprint. This includes up to 30 percent performance improvement in PPS handling compared to the earlier Intel processor.

AT&T’s vCDN (virtual Content Distribution Network) is a service that allows customers to manage and distribute video and multi-media web content across networks.

“We saw 25 to 30 percent performance improvements from moving it (vCDN) to Purley,” Rice said, referring to the Intel processors’ code-name. “It was a pretty seamless transition, moving it from the older Intel CPUs onto the new one. We are able to do more with fewer processors, and we’re able to get more capabilities out of our CDN and grow it horizontally as well.

“And all of the improvements, whether on the process side or the architecture side, they all have some networking improvement piece as well,” Rice added.



These performance improvements are helping AT&T move closer toward its goal of virtualizing 75 percent of its network by 2020. During its second-quarter earnings call last month, AT&T CFO John Stephens told investors that the company has virtualized more than 40 percent of its network functions. It’s making progress toward its network functions virtualization goal of 55 percent by year-end.

“We want to make sure the whole ecosystem moved with us toward network virtualization,” Rice said. “We didn’t want to have something special just for AT&T. We wanted it to be for the whole industry.”

Additionally, achieving network performance improvements requires automation, Rice said. “You’ll never get to those percentages without automation being a key part, he added.

In an earlier interview with UBB2020, Rice said:

“As we move down an automation path, as we move down a machine-learning path to drive more automation, [having open interfaces on network elements] is really a necessary first step — these open interfaces that cannot be skipped over or overlooked. I don’t know that people understand the significance of that.”



AT&T, Verizon, Spectrum Enterprise lead in fiber lit business connections; M&A in 2017

Vertical Systems Group (VSG) ranked the leading providers of on-net fiber business connections as of the end of 2016. The research group said that retail and wholesale fiber providers with 10,000 or more on-net fiber-lit commercial buildings in the U.S. qualify for this new benchmark.

VSG’s 2016 U.S. Fiber Lit Buildings LEADERBOARD list includes a mixture of traditional telcos, cable providers and a competitive carrier: AT&T, Verizon, Spectrum Enterprise were the top three followed by CenturyLink, Comcast, Level 3, Cox, Lightower Fiber NetworksZayo, Altice USA and Frontier.

The Challenge Tier of fiber providers includes companies with lit fiber connections to between 2,000 and 9,999 U.S. commercial buildings. Seventeen companies qualified for the 2016 Fiber Lit Buildings Challenge Tier as follows (in alphabetical order): Cincinnati Bell, Cleareon, Cogent, Consolidated Communications, Electric Lightwave, Fairpoint, FiberLight, FiberNet Direct, FirstLight, IFN, Lumos, Southern Light, Sunesys, Unite Private Networks, Uniti Fiber, Windstream and XO.


Rosemary Cochran, principal at Vertical Systems Group, said the criteria for the leader board list was for network service providers to have fiber installed and fiber transport equipment ready to serve business customers.

“This is commercial buildings and data centers that have fiber in place and there is active service equipment that enables provisioning of commercial services,” Cochran said. “We’re not counting residential fiber or standalone cell towers.”

Cochran added that VSG is not counting near-net buildings where service providers may be passing buildings with fiber but have not connected them yet.  “It is either lit or not lit,” she said.

“On-net fiber lit buildings are valued strategic assets that give retail and wholesale providers a competitive edge in profitably delivering services to business customers. A major benefit of a fiber lit building is ready connectivity with provisioning through service orchestration, without the construction cost and extensive lead time required to light a building,” Cochran added.


Acquisitions of Fiber Providers happening this year:

“These dynamics are driving this year’s acquisitions among fiber providers that will significantly impact the U.S. fiber landscape. Eighteen of the twenty-eight Fiber LEADERBOARD and Challenge Tier companies have fiber-related transactions just completed or pending.”

One of the largest out of this group is CenturyLink’s pending deal for Level 3, one that will enhance the telco’s on-net fiber footprint.

CenturyLink’s Level 3 acquisition will increase its reach by nearly 75% to approximately 75,000, including 10,000 buildings in EMEA and Latin America, giving the telco a larger footprint to deliver Ethernet and software-defined services.

Some of the other acquisitions that will alter the on-net fiber profile will be Verizon’s recently completed purchase of XO Communications and Crown Castle’s pending acquisition of Lightower.

By purchasing XO, Verizon gained metro fiber networks in 40 major U.S. markets with over 4,000 on-net buildings and 1.2 million fiber miles.

Cochran said that since a number of these deals have not been completed it remains unclear as to what effect they will have.

Crown Castle, which will gain an additional 22,000 buildings, stands out from the crowd since the service provider has been operating the fiber providers it has bought as separate companies.

However, other pending acquisitions being made by Consolidated Communications, Cincinnati Bell and newer players like Uniti Fiber will have an effect.

Consolidated Communications, which will announce its second-quarter earnings tomorrow, recently completed its acquisition of FairPoint. By acquiring FairPoint, Consolidated immediately established itself as the ninth largest fiber player with a presence in 24 states and 8,000 on-net buildings.  This greater density will enable Consolidated to pursue more dark fiber and lit Ethernet service opportunities with a larger mix of business and wholesale customers.

“Out of the companies on the leader board, 18 of them either already completed acquisitions or some are pending,” Cochran said. “Because a lot of deals happened in July with some that are pending such CenturyLink/Level 3, they are not reflected here.”


Strategic importance of Fiber Assets to deliver business services:

Regardless of when these deals are completed, Cochran added that they show how having a large arsenal of fiber is important to compete for business services.

“There’s a lot that happened in the first half of this year and that’s going to shake things up,” Cochran said. “Just the fact that you have more than half of these companies involved in some kind of transaction shows the value of fiber.”

On the fiber leaderboard, 7 service providers have an ongoing spot on VSG’s Carrier Ethernet Leaderboard report. These providers include: AT&T, Verizon,Spectrum Enterprise, CenturyLink, Comcast, Level 3, and Cox.

Cochran noted that the presence of fiber has coincided with the growth of Ethernet in the domestic U.S. market.

“Fiber-based Ethernet is the most widely deployed technology,” Cochran said. “As those upgrades take place, one of the drivers is trying to get more bandwidth so there’s a correlation between higher bandwidth services and having that fiber in the building.”

“Wholesale services are also a big driver, which becomes a question of service provisioning.  The work that’s being done at the MEF on inter-carrier provisioning and service orchestration across carriers to automate it with standard APIs would help to accelerate new services,” she added.


Other Fiber Providers:

All other fiber providers with fewer than 2,000 U.S. commercial fiber lit buildings are in the Market Players tier. The 2016 Market Players tier includes more than two hundred metro, regional and other fiber providers, including the following companies (in alphabetical order): Alpheus, Axia, Birch, C Spire, Centracom, Conterra, DQE Communications, EarthLink Business, Fatbeam, Global Capacity, GTT, Hawaiian Telecom, Hibernia, Hunter Communications, Independents Fiber Network, Infostructure, Logix Communications, LS Networks, Mediacom, Monmouth Telecom, Orca Communications, Pilot Fiber, PS Lightwave, Shentel, Silver Star Telecom, Sovernet, Spirit/Palmettonet, Syringa, TDS Telecom, TPX Communications, U.S. Signal, Veracity, Wilcon, WOW and others.


Comment & Analysis of Century Link’s 2nd Quarter results

Century Link today reported second quarter 2017 results.  The company:
– Achieved second quarter operating revenues of approximately $4.1 billion
– Generated operating income of $367 million in second quarter, which reflects approximately $150 million of one-time charges related to the sale of the data centers and co-location business on May 1, 2017 (Co-location Sale)
– Generated adjusted EBITDA of $1.44 billion in second quarter, excluding special items
– Revenue for the quarter fell 7.0% to $4.09 billion, which was down from $4.40 billion last year.
– Continued to invest to drive higher broadband speeds throughout its network footprint.
– Ended the quarter with more than 3.8 million addressable units capable of speeds of 100M bits/s or higher and more than 1.5 million addressable units capable of 1G bits/s or higher
– Continue to anticipate completion of the acquisition of Level 3 Communications by end of September 2017
Author’s Comment: 
Negative surprises (see Presentation section below):
1. Year over year (y/y) decline in Carrier Ethernet revenue, yet solid growth in (older) MPLS revenue is quite puzzling.
2.  10.1% decline in enterprise legacy services revenues due to lower voice and low-bandwidth data services revenues.
3. Fewer consumer broadband subscribers in past year. Strategic consumer revenue decreased 4.0% Y/Y – impact of satellite contract restructuring in 1Q and lower
broadband units. Consumer legacy revenue declined 8.8% Y/Y – lower access lines.
Positive surprises:
1. Enterprise strategic revenues (not including co-lo sales) grew 4% while high-bandwidth data services revenues increased 5% Y/Y.
2. Operating expenses for enterprise segment declined $87 million, or 6.3%, Y/Y – sale of co-location business reduced operating expenses ~$60 million; reduction
in employee-related expenses.
3.  Operating expenses for consumer segment declined $47 million, or 7.4%, Y/Y – lower employee-related expenses.
Most astonishing to me was that the Century Link executives did not even mention their public cloud computing business, which was heavily promoted in 2011 when the company acquired Savvis.  We assume the company’s entertainment video business results were included in the consumer segment, but they were not separately disclosed.
Century Link CEO Comments:

“We are confident our continued investment in high-quality, high-bandwidth broadband network infrastructure positions Century Link well for long-term growth,” said Glen F. Post, III, Century Link chief executive officer and president.

“Enterprise demand for high-bandwidth data services remains strong and, while consumer broadband units were weaker than expected, we are encouraged by the higher-value customers our improved offerings are attracting. We accelerated our capital investment in high-bandwidth services and broadband infrastructure during the second quarter, which we believe better positions us to increase revenues in the second half of 2017 and beyond. We anticipate second half and full year 2017 capital expenditures of approximately $1 billion and $2.6 billion, respectively.

“We achieved our expected adjusted EBITDA for the quarter as our employees did a great job managing costs, while core revenues were below our expectations primarily due to the decline in legacy revenues and the decline in broadband units being higher than anticipated. We continue to make good progress in obtaining the necessary approvals for the pending Level 3 acquisition, having received clearance in 23 of 25 required states and territories. Integration planning is progressing well and we continue to anticipate completing the acquisition by the end of September 2017. We remain excited about the value we believe this transaction will create for our customers, our shareholders and our employees,” concluded Mr. Post.


Century Link’s presentation – 2ndQ 2017 Highlights & Trends:

1.  Strategic services:
◦ Enterprise high-bandwidth data – solid growth in MPLS revenue offset by decline in Ethernet revenue; grew 5% Y/Y on a normalized basis
◦ IT & Managed services – continued growth in IT Services along with stabilization of managed hosting revenue
◦ Enterprise other strategic – ~$100 million Y/Y and Q/Q decline due to Colo-cation sale
◦ Consumer broadband – fewer subscribers Y/Y
◦ Consumer video – impact of satellite video contract restructuring

2.  Acquisition of Level 3 Communications:
◦ Continued good progress in obtaining necessary approvals; received shareholder
approval and regulatory approvals/clearances in 23 states or territories, with 2 states
◦ Integration planning process continues to go well; remain confident with cash synergy target of $975 million
◦ Named combined company senior leadership team, effective at close; overall
organization design progressing well
◦ Continue to anticipate closing by end of September 2017

Forward Guidance:

Based on first half 2017 results and current expectations for the remainder of the year, Century Link anticipates coming in slightly below its full-year 2017 revenue and adjusted diluted EPS guidance, primarily driven by higher legacy revenue declines and lower consumer broadband revenue growth than anticipated. The company continues to expect adjusted EBITDA and adjusted free cash flow to be near the lower end of prior guidance.

Century Link is not providing updated guidance ranges for full-year 2017 due to the pending acquisition of Level 3, currently anticipated to be completed by the end of third quarter of 2017, and the expected consolidation of results for the combined companies in fourth quarter 2017.

Earnings Call:

The call will be accessible for replay through August 9, 2017, by dialing 855-859-2056. Investors can also listen to Century Link’s earnings conference call and webcast replay by accessing the Investor Relations portion of the company’s website at through August 24, 2017.   Here are a few highlights:

First, we continue to outperform MPLS market growth projections forecasted by leading industry analysts. In second quarter, we had nearly 2,000 MPLS customers. This performance was driven in particular by our SMB customers, where we are seeing improved install intervals of nearly 20%, which should help accelerate our revenue recognition as we move into the third and fourth quarters of this year. Next, we launched a number of new products and – we’ve launched a number of new products in the past few months including our CenturyLink Ethernet service.

We’ve also had three simplified bundles of SD-WAN plus network packages. We rolled out a competitively priced cloud-enabled small business VoIP offering. And we rolled out a new comprehensive managed Enterprise offering that is an end-to-end solution that includes WiFi management, network management, video surveillance, security and mobility management, all from a single interface. Also we have increased focus on customer retention and we are seeing lower credits and adjustments as a result.

In addition, CenturyLink continues to be one of the leaders in network virtualization through the deployment of software-defined networking and network virtualization capabilities. Based on initial results, we expect these services to create significant value in the months ahead. Also the continuous onslaught of new security threats, such as WannaCry has brought greater interest in and sales of our strong network and cyber security capabilities, as we believe CenturyLink is growing in recognition as a leading provider of security services that are so important to our Enterprise customers.

And lastly, based on third-party research support, U.S. Enterprise high-bandwidth data services are forecast to grow at mid-single-digit compounded annual growth rates through 2021, and U.S. Enterprise Managed Network Services are forecast to grow at mid to upper single-digit compounded annual growth rates through 2021. Now this forecast gives even more confidence in the opportunity to continue to grow Enterprise business in the months ahead.

Second, our IT services revenue, which is primarily driven by IT consulting, cyber security, IT service management and big data and analytics, is growing. And our managed hosting business also showed a solid turnaround this quarter. The team overcame the market confusion and sales disruption created by the colocation sale and grew cloud revenue, especially driven or aided by our Cloud Application Manager suite.

Next as expected, we had a seasonally challenging quarter from consumer broadband subscribers approximately 65,000 residential subscriber loss was higher than anticipated. This was driven to a great degree of our stronger cable competition, particularly 1 gig offerings in some of our key markets, coupled with aggressive pricing. Over the past year, we have made a pivot towards higher-quality, more profitable consumer broadband sales by removing several low-priced promotional offers and increased credit standards.



CenturyLink Delivers Broadband Services on CORD Platform

J.D. Power: SMB a Growth Opportunity; Telecom ARPU Falling in Every Region