F5G-A All-Optical ADN: Unlocking a New Era of Premium Operations with AI for Networks

Carriers of optical networks used to benefit from a vast user base and that base's requirements for higher bandwidth. However, the landscape has changed with the rise of AI, where users demand more than network speeds. They expect deterministic network experience for AI applications featuring an immersive experience and device-cloud synergy. These requirements are driving carriers to advance into a new era of operations – the era of premium operations.

The need for such an evolution can be understood from two perspectives: the household perspective and the network perspective.

The household perspective: Intelligent home appliances and AI applications are increasingly becoming part of our everyday lives. An expanding array of whole-house intelligent devices need to connect to networks to make lifestyles more efficient and convenient. Within the next five years, the average household is expected to have more than 50 networked devices. Furthermore, tens of millions of applications already exist for gaming, livestreaming, and education, with a range of new AI applications in areas like text generation, digital art, and knowledge interaction constantly emerging.

Driven by a clear need to better adapt to this intelligent upgrade of home devices and applications, fiber to the room (FTTR) deployment is rapidly gaining popularity. As of today, there were already 50 million FTTR users worldwide, and this number is expected to exceed 100 million over the next three years. With the proliferation of networked intelligent devices, rapid expansion of AI applications, and growth of home networking, carriers must upgrade their experience assurance capabilities to meet the evolving household demands for a seamless, responsive, and differentiated network experience.

The network perspective: In the AI era, networks must be able to efficiently connect services to computing centers, and connect computing centers themselves, all while ensuring fast and stable connectivity. For instance, AI-powered industrial robots require a network latency of 5 ms and less, while cloud phones and cloud PCs demand latency of less than 10 ms. Intelligent computing centers also require more resilient connectivity than traditional DC interconnect (DCI) in order to support collaborative training. A link disruption of more than 10 seconds means training will need to be rolled back and start all over again, resulting in a potential million-dollar loss.

Optical networks serve as the fundamental infrastructure to address all these requirements for smart homes and computing center interconnection. To ensure exceptional user experience and enhance operational efficiency, these networks must undergo comprehensive quality-oriented upgrades.

At the heart of this transformation is the optical network operations platform, a vital part of F5G-A technology that focuses on intelligence. Huawei's all-optical autonomous driving network (ADN) platform is continuing to evolve using AI foundation models and digital twins, enhancing both user experience and operational efficiency in two critical areas: premium broadband and premium transmission networks.

FAN Spirit (Huawei home broadband agent): Intelligent operations for premium broadband

Experience monetization is an effective way for carriers to build a stronger bond with users and effectively improve their networks. A primary challenge, however, exists for carriers because the user experience is usually intangible and unperceivable, leading to a number of issues. For example, they identified potential users mostly through making widespread sales calls without clear targets, provided experience assurance only after users have filed complaints, and troubleshot their networks manually.

Carriers have made numerous attempts to improve their experience monetization methods. One carrier in Argentina set up an over 500-strong marketing team to identify potential installed-base users they could recommend suitable upgrade packages to, but only 3% of their sales attempts were successful. Another carrier in China previously had a team of more than 4,500 employees across its customer service center, business support, installation, and comprehensive scheduling departments tasked with handling complaints from high-value users. However, they still suffered from an over 10% churn rate. Examples like this show that a reactive, complaint-driven operations approach is both costly and ineffective.

Building an intelligent O&M agent that can proactively detect problems before they occur and automatically fix the issues that impact user experience is therefore imperative for carriers. These agents can efficiently guarantee home broadband (HBB) experience. Huawei FAN Spirit is one such agent that can serve as an AI manager for networks thanks to two innovative features: its customer experience index (CEI) measurement system and the HBB foundation model.

The CEI experience measurement system can be used to tackle the intangibility of user experience by quantifying the experiences of individual subscribers. With it, experience indicators are first collected from network elements (NEs), networks, and applications to build a comprehensive experience evaluation system that covers eight dimensions. Then, intelligent algorithms are used to determine the correlation between each dimension and user experience. These dimensions are given initial weights from which a score is calculated. Throughout this process, the algorithms also automatically optimize weighting to further improve scoring accuracy.

The HBB foundation model serves as the "brain" that comprehensively manages HBB operations. It can automatically analyze various issues during HBB experience monetization and provide related solutions. The model features three highlights. First, it has been trained with the industry's most comprehensive expert experience library and a corpus containing over 100 billion entries to cover over 99% of HBB operations expertise and scenarios. Second, it integrates HBB fault graphs and chain of thought (CoT) inference algorithms to deliver industry-leading speeds for issue analysis and troubleshooting. Third, it is being continuously further trained and optimized with updated experience libraries to cover more scenarios and become more efficient on the go.

Through these two innovative features, the Huawei FAN Spirit benefits carriers in HBB user development, experience optimization, and complaint handling.

Intelligent potential user identification for faster HBB business development: Most carriers have traditionally identified potential users using two indicators: service packages and NE KPIs. This approach lacks the refined granularity needed to accurately reflect users' experience and results in low marketing success rates. The Huawei HBB Agent expands the scope of user profile tagging to cover more than 110 tags that together create 6 indicators: a customer experience index (CEI), Wi-Fi coverage, home stations, user activeness, application preferences, and home networking. This allows carriers to more comprehensively and accurately identify potential users. One carrier in the Middle East has already used the FAN Spirit to boost their marketing success rate by over six times and reduce their customer acquisition cost by 32%.

Intelligent experience optimization for proactive HBB experience assurance: Traditionally, carriers' HBB experience operations are mainly driven by user complaints, resulting in high complaint and churn rates. With FAN AssurSpirit (FAN Spirit for proactive assurance scenario), carriers can take a more proactive approach. The CEI scoring model detects and measures issues related to poor quality of experience (QoE), which are then automatically diagnosed by the HBB foundation model and addressed with rectification solutions. Poor-QoE issues that can be remotely optimized are automatically processed and resolved. When it comes to poor-QoE issues that require on-site rectification, work orders are automatically dispatched to field engineers for proactive on-site visits. This way, carriers can provide proactive assurance by resolving issues before they lead to complaints. A provincial branch of China Mobile has used the FAN AssurSpirit to reduce complaint-related work orders by 60% and cut churn rate by about 20%.

Intelligent fault diagnosis for automatic handling of HBB complaints: Currently, carriers primarily rely on site visits by field engineers for HBB fault diagnosis. The troubleshooting process requires highly experienced engineers and involves complex manual information queries. The FAN CompSpirit (FAN Spirit for complain handling scenario) offers the ability to automatically handle fault complaints, halving both the number and duration of site visits by field engineers. The FAN CompSpirit provides a fault diagnosis system for automatic remote fault locating that uses diagnosis templates distilled from actual experience handling the 30+ most common user-reported faults. The FAN CompSpirit also provides a troubleshooting assistant for field engineers, allowing them to query information about more than 30 troubleshooting scenarios without needing to consult backend personnel. A provincial branch of China Mobile has used the FAN CompSpirit to reduce the average troubleshooting duration by 53%, the number of site visits by 50%, and the number of work orders requiring backend support by 59%.

Transmission Digital Map: Zero-outage networks with deterministic reliability

AI services need networks that provide higher reliability and more resilient connectivity. Traditional network planning mainly relies on protection switching for reliability assurance. However, optical fibers are dumb resources and incapable of active processing or management. This means most faults on optical networks are invisible and difficult to manage.

Multiple investigations have revealed widespread structural risks across optical networks, such as single-node and co-cable networking, alongside fiber degradation and intermittent disconnections. These risks can lead to protection switching failures and service interruptions. For instance, a fault at a carrier’s site in Italy resulted in a 24-hour interruption of 7.6 Tbit/s of services and significant financial losses. Moreover, conventional protection switching cannot ensure lossless services and will fail in the case of two fiber cuts or more. AI computing centers, for example, require highly resilient connections – if a link is interrupted for more than 10 seconds, AI training will roll back, leading to million-dollar losses.

At the same time, traditional fault maintenance involves lengthy troubleshooting cycles that can last several hours and require at least three teams to inspect outdoor fiber routes, equipment rooms, and user-side facilities in order to locate a fault. This extended troubleshooting significantly increases operational expenditure (OPEX). For example, in the Philippines, a carrier's network experiences over 1,000 fiber cuts annually, with an average recovery time exceeding 24 hours, which severely impacts service availability.

For carriers, the primary challenge in improving network reliability lies in managing "invisible and unmanageable" resources, especially dumb resources like the fiber and transmission paths in optical networks. To address this challenge, Huawei provides the NCE-T Transmission Digital Map solution, which leverages digital twin technologies to realize digitization, visualization, and holistic awareness for transmission networks. It offers real-time risk detection, instant resource queries, and automatic fault locating – all presented through an intuitive, single-pane-of-glass interface. Building on this map, an E2E reliability system can be established, integrating prevention, self-healing, and restoration mechanisms. These innovations empower carriers to upgrade their networks to a model that guarantees seamless, resilient connectivity even in the face of fiber cuts.

Transmission Digital Twin: Optical networks comprise numerous dumb resources such as optical cables, transmission paths, and wave channels. The locations, distances, and loss parameters of these resources cannot be automatically detected, making it extremely difficult for carriers to acquire network information in real time or in full detail. The Transmission Digital Twin addresses this challenge by digitizing and visualizing transmission network resources across four layers – fiber, optical, electrical, and service – and three service dimensions – bandwidth, latency, and availability. Specifically, leveraging intelligent devices such as Fiber Doctor, Light Sensor, and BianQue AI Optical Module, the system collects network data 24/7 from the fibers, devices in equipment rooms, and user-side facilities. Additionally, these intelligent devices collect samples every few milliseconds and locate faults with meter-level accuracy. This allows carriers to keep track of network and resource changes continuously in real time.

The Transmission Digital Twin performs holographic perception of network changes. This full-factor, 360-degree method enables carriers to establish an E2E reliability system with the following functions:

Proactive prevention: Traditional O&M methods struggle to effectively identify latent network faults and determine potentially impacted services. NCE-T Transmission Digital Map provides an OTN AssurSpirit to address this issue. This agent leverages algorithms such as multimodal deep learning to visualize more than 42 types of network risks in real time. Moreover, by integrating intelligent hardware such as the Fiber Doctor and distributed acoustic sensing (DAS), the agent automatically pinpoints the root causes of network risks to proactively resolve them and prevent failures before they occur. One carrier in China using the OTN AssurSpirit to support a dedicated network carrying government and enterprise services has significantly enhanced network resilience, reducing fault-related work orders by 40%.

Dynamic resilience: When a failure occurs, traditional protection switching takes up to 50 ms and is extremely likely to cause a disruption to service continuity. Using innovative DC-OTN technology, NCE-T Transmission Digital Map provides seamless protection switching with zero packet loss and ensures an unaffected user experience. Additionally, its advanced ASON capabilities can reroute services within 50 ms, enabling service self-healing even in the event of multiple network failures.

Rapid healing: A single network fault can trigger many associated alarms, some of which are often repeated or false alarms. When this happens, the NOC center will issue a large number of work orders. Traditionally, O&M engineers need to sift through an ocean of alarms to quickly locate the fault. Three teams need to be dispatched to the cables, equipment room, and customer premises for on-site troubleshooting, which is both time- and labor-intensive. NCE-T Transmission Digital Map comes with an OTN FaultSpirit that uses a failure propagation graph algorithm to automatically identify the root causes of faults. It also uses intelligent hardware such as the BianQue AI Optical Module or embedded optical time domain reflectometer (eOTDR) to quickly demarcate and locate faults. One provincial branch of China Mobile in southern China has used this OTN FaultSpirit to slash their average fault locating time from 4 hours to 15 minutes and reduce their number of on-site teams per fault from 3 to 1, creating significant savings in both troubleshooting manpower and OPEX.

The demand for home networking and intelligent applications is growing rapidly, and the need for rapid access to compute and efficient interconnection between computing centers is increasing at pace. These demands represent huge development opportunities for carriers. Huawei is integrating AI into optical networks and helping carriers build L4 Autonomous Networks (AN). These networks will better position carriers to intelligently and efficiently develop new services, guarantee user experience, and maintain their networks, while also giving them more opportunities to grow revenue and reduce costs. Huawei is ready to work alongside carriers to embrace premium operations in the F5G-A all-optical era and unlock new potential with AI.