Accelerating Intelligent ICT with a Four-layer Framework and Transformation in Three Areas

Rapidly evolving AI technology is bringing unprecedented opportunities and transformation potential to telecom carriers. At the business level, AI applications can help carriers increase revenue by developing smart and easy-to-use services. For telecom operations, AI can also help carriers manage and optimize network resources better and improve O&M efficiency. At the same time, carriers need to explore and apply ICT solutions to support AI by providing stronger data connection, storage, transmission, and computing capabilities.

Centering around data, Huawei has proposed a four-layer framework and transformation in three areas for the intelligent era (Figure 1). The four layers are service, network management, cloud infrastructure, and network infrastructure. The three areas of transformation are business, experience, and infrastructure. Intelligent technology has changed the way data is produced, managed, applied, stored, processed, and transmitted on the four layers, and carriers need to adapt to transformed business models, O&M-based experience, and cloud-network infrastructure in the intelligent era.

Business transformation: A new chapter for AIGC and business value

New ways of producing data are driving business transformation. For example, AI is learning to create new worlds through AI-generated content (AIGC). This will potentially take AI to a new level—artificial general intelligence (AGI)—which will drive transitions to new business models. In the text-to-image and text-to-video fields, foundation model algorithms and applications are maturing, with diffusion-model-based algorithms already being widely used for image generation. The Transformer architecture used by text-to-video models like Sora follows scaling laws and realizes video generation abilities.

AIGC is dramatically transforming content generation and seeing fast adoption in fields like picture generation, short video creation, and movie production. AIGC is also transforming connectivity from connections between people to comprehensive connections between people, objects, digital humans, smart vehicles, and drones. These new connections involve not only manual operations, but automation and machine-to-machine communications. In addition, AIGC is transforming interactive experiences by combining AI, 3D, VR, and AR technologies to enable a leap from content creation to holographic interactions.

Business (or business cycle) transformation means new ways of producing data are increasing content production efficiency (machines, not just humans, as producers), content complexity (HD, 3D, and holographic content) and distribution speeds (faster interactions with and ubiquitous access to content). This is generating massive amounts of data and connections. It is predicted that by 2030, new digital content will increase network traffic by over 10 times and increase the number of connections from tens of billions to hundreds of billions, meaning new business opportunities for carriers. New ways of producing and transmitting data are also transforming business models to enable innovation in personalized products and services and improve user experience and loyalty.

Experience transformation: A future with intelligent network O&M

In the field of network O&M, intelligent technologies are changing service rollout, resource configuration, and O&M processes. These all center on transforming user experience. The intelligent transformation of network infrastructure will follow three trends:

First, from event-driven networks to intent-driven networks. Traditional event-driven networks respond only to specific events, while intent-driven networks can understand user intent and adjust network configurations to user needs rapidly and efficiently.

Second, from empirical decision-making to AI-assisted decision-making. Traditional network management decisions are made based on expert experience and analyses. With the help of AI, network management personnel can make decisions more efficiently using AI algorithms and data analytics, which also improves network performance and reliability.

Third, problem-solving evolution from scenario-specific models to foundation model generalizability. Traditional AI models are created to solve specific problems in specific scenarios. Foundation models can be used to create more generalizable problem-solving systems. When used alongside scenario-specific models, they can solve complex problems through analyses that streamline different network scenarios and processes.

Telecom networks use the following three layers of solutions to make network O&M intelligent:

• Network element (NE) intelligence:

  More real-time sensing components and AI inference capabilities can be added to network equipment. AI-native hardware will enable networks to be capable of finer-grained sensing and real-time synchronization.

• Single-domain intelligence:

  An intelligent management, control, and analysis platform can be used to create digital models of networks that associate discrete data related to network resources, services, and status, and provide digital twins adapted to different domains. This platform's single-domain, high autonomy capabilities, from data collection, sensing, and analysis to simulation, decision-making, and control, can ensure guaranteed network connection quality and timeliness. For example, Huawei provides an IP-based digital map with six visualized layers—physical, network, slicing, routing, service, and application—as well as capabilities such as congestion view, experience view, and fault view.

• Cross-domain intelligence:

  Collaboration can be performed across domains. Huawei provides intelligent platforms such as NCE-Super, ADO, and SmartCare to support collaboration across domains, such as fault demarcation through coordination between IP and optical domains, and intelligent orchestration for cross-domain services.

Telecom foundation models will be key to realizing intelligent network O&M. Carriers' complex O&M processes in all manner of scenarios across domains will require the capabilities of telecom foundation models and single-domain autonomy to implement cross-domain collaboration and the autonomous completion of complex processes. Huawei's Telecom Foundation Model, for example, consists of three layers: the foundation model, telecom industry models, and applications. The foundation model layer provides multimodal, computer vision, natural language processing (NLP), scientific computing, and prediction based on Huawei's Pangu models and third-party open-source foundation models. The telecom industry model layer provides industry-specific capabilities using high-quality industry corpuses and efficient toolchains. The application layer provides role-based copilots and scenario-based agents to improve employee efficiency and customer satisfaction. Huawei's Telecom Foundation Model provides out-of-the-box applications and supports local deployment and incremental training by carriers to deliver intelligent technologies that meet different user needs.

O&M experience transformation uses intelligent technologies like foundation models, and the three-layer intelligent telecom network solutions to bring intelligence to network infrastructure and build AI-native, intent-driven, and digital twin capabilities. These features can help carriers guarantee user experience, agilely roll out services, optimize resource configuration, and improve O&M efficiency, accelerating network evolution towards autonomy.

Infrastructure transformation: The network foundation for the AI computing era

The nascent AI computing era is seeing a major leap in computing technology from general-purpose computing to AI computing. This means the number of foundation model parameters and demand for computing power will grow significantly. Trillion-parameter foundation models will require AI computing centers that deliver 10,000-GPU computing power and efficient data management. This is making coordination across computing, storage, networks, management, and efficiency a key trend. Networks that support AI service growth are critical to realizing ubiquitous computing power and efficient data flow. End-to-end collaboration between devices, edge, and cloud across all manner of scenarios will improve overall computing efficiency.

Two technologies are key to achieving network-cloud synergy.

The first is hyper-converged DCNs. AI computing is characterized by small data flows with large packets, which can easily lead to load imbalance, decreasing network throughput and reducing computing efficiency. Hyper-converged Ethernet DCNs can meet data center service requirements in different phases of development and different scenarios. Ethernets that integrate general-purpose computing, storage, HPC, and AI improve deployment and O&M efficiency and reduce maintenance costs. Huawei's innovative network scale load balancing (NSLB) algorithm, for example, optimizes traffic paths to achieve global load balancing, delivering 20% higher Ethernet efficiency than the industry average and 10% higher than IB.

The second technology is elastic data center interconnect (DCI) and data center access (DCA) networks. AI computing requires fast transmission of massive amounts of data for training and inference. Traditional private lines are constrained by their lack of service elasticity and high costs. Superior to the current common practice of shipping hard drives via express delivery, Huawei's innovative elastic transport networks deliver end-to-end 400GE, SRv6, and elastic high throughput. These networks use AI-powered data flow identification and path orchestration, as well as high throughput and high-reliability elastic 10GE private lines, to provide users with elastic, task-based data express services (same-hour, same-day, or next-day delivery).

Traditional cloud infrastructure urgently needs upgrading in areas such as data center power supply and cooling, hardware infrastructure clustering, and the integrated deployment and collaborative O&M of computing, storage, and network resources. Key requirements for network infrastructure in the cloud and intelligence era include ultra-high bandwidth and ultra-low latency, moving storage and computing from devices to the edge, high security and reliability, and task-based network services.

The four-layer framework and three-area transformation were proposed to systematically outline the carrier opportunities presented by the intelligent era and help carriers build the required capabilities to seize these opportunities. The convergence of networks, cloud, and intelligence is already a major trend, as new opportunities are created by intelligence powering ICT, and demand for infrastructure upgrades is generated for ICT to boost intelligence. Huawei looks forward to working with customers and partners in exploring new architectures, applying new solutions, and jointly developing new technologies to reap the value of intelligence for data and traffic.