Eight Innovations in China Unicom’s 2025 Network Architecture

In March, 2021, China Unicom launched a sequence of white papers on next-generation digital infrastructure, or CUBE-Net 3.0, and arrange an indication base in the Guangdong–Hong Kong–Macao Greater Bay Area to advertise innovation.

Key improvements of CUBE-Net 3.0

The CUBE-Net 3.0 technical structure outlined by China Unicom (Figure 1) consists of eight elements:

China Unicom,Network Architecture

CUBE-Net 3.0 technical structure

1. Ubiquitous and elastic cell broadband

With the regular growth of 5G networks and purposes, 5G companies have gotten more and more differentiated, convergent, and diversified. However, uplink bandwidth stays a bottleneck. 

Mobile broadband communications have to evolve in direction of increased spectrum bands and bandwidth. Due to the large quantities of connections, numerous low- and medium-rate companies will coexist with companies that require massive bandwidth for a very long time. As cell broadband requires extra clever, versatile, and elastic bandwidth administration, it’ll develop into ubiquitous, elastic, clever, and inexperienced over the following decade.

Flexible bandwidth: Full-spectrum 5G reconstruction will lay the muse for the continual evolution and growth of 5G networks. To obtain that, three main necessities have to be fulfilled: First, driving evolution in direction of working bandwidth above 1 GHz with massive single-carrier bandwidth above 400 MHz in millimeter wave and terahertz bands. Second, creating spectrum useful resource swimming pools the place cross-band assets are managed in swimming pools by coordinated on-demand entry to high-, medium-, and low-frequency bands. Third, matching bandwidth assets to customers’ service calls for by clever bandwidth useful resource allocation.

Flexible timeslots: High uplink and downlink calls for for B2B and B2C companies are utterly totally different. Alongside interference avoidance, versatile timeslot configuration is required for big uplink bandwidth to attain optimum useful resource utilization and person expertise. AI can precisely predict service demand traits and thus decide the required timeslot configuration. Real-time changes to the timeslot configuration based mostly on service demand are required to attain symbol-level versatile configuration. We can use AI to establish interference options and stop interference to match companies with capabilities.

Dedicated good uplink: Built on good timeslots, service aggregation, and devoted units, devoted good uplink works with AI to spice up time, frequency, energy, and antenna concurrency. It additionally opens up the chance for future networks to supply gigabit uplink.

Wireless AI: 5G and AI can precisely match radio assets to customers with converged scheduling algorithms based mostly on service demand, person grouping, scheduling prediction, and community MCS/RANK high quality. Technologies for community, person, and repair sensing can perceive service intentions and intelligently choose and coordinate assets between networks. This contributes to optimized mobility administration, interference administration, and cargo balancing to attain clever service navigation.

2. Ultra-broadband, high-quality optical fiber entry

Fixed entry networks are evolving from 1G PON and Wi-Fi 5 to 10G PON and Wi-Fi 6, offering gigabit entry to properties. In the following 5 to 10 years, optical entry networks will develop in the next three instructions.

Direction 1: extending all-optical entry to the top. OLTs might want to help a number of types of versatile networking for the varied companies of the longer term. Therefore, future optical fibers will proceed to be prolonged to the top. Gigabit FTTR will probably be deployed in rooms to attain steady entry with excessive bandwidth and low latency.

Direction 2: offering differentiated slice bearing and open capabilities. OLTs will help end-to-end slicing to fulfill the differentiated bearing necessities of companies resembling house broadband, authorities and enterprise companies, and {industry} purposes. The uplink of OLTs helps VXLAN, VLAN, ODU/OSU, SRv6, and different forwarding and routing choice modes. Adopting applied sciences resembling HQoS, arduous isolation in chips and PON ports, and Wi-Fi air interface slicing, OLTs work along with bearer community slices to allow end-to-end slicing. Traffic is directed to totally different slices based mostly on companies and purposes to create larger synergies with service capabilities on the cloud.

Direction 3: opening up embedded computing energy. Computing assets deployed on units and the sting may help obtain software sensing and real-time pipe optimization. AI-enabled cloud-edge-device collaboration helps software eventualities resembling expertise administration, figuring out potential clients, and clever O&M. Video optimization, video surveillance backhaul, and industrial IoT may be migrated to cloud by opening up the embedded computing capabilities of OLTs and ONTs and collaborating with the cloud ecosystem. 

3. Smart and open all-optical base

The CUBE-Net 3.0 all-optical base has 4 key options: (1) Stable structure, all-optical switching, massive bandwidth, and a excessive degree of scalability. (2) All-optical anchor and all-service entry help one-hop entry to the cloud. (3) Intelligent administration and management for computerized O&M. (4) End-to-end slicing that gives differentiated SLA companies for a number of companies on the identical community.

The CUBE-Net 3.0 all-optical base additionally adopts 4 key applied sciences:

1. All-optical cross-connection (OXC). OXC gear that adopts wavelength selective switches and optical backplanes can obtain zero fiber connections, plug-and-play, versatile scheduling, easy capability growth, and ultra-large-capacity wavelength grooming, enormously saving gear room house and lowering energy consumption.

2. Optical service models (OSUs). Service-oriented OSUs can effectively transport low-rate companies and supply a slicing answer for the all-optical base, representing a possible evolution route for OTN applied sciences.

3. Centralized and distributed management aircraft protocol structure. Centralized SDNs use international assets to compute optimum service paths. Distributed management protocols allow agile and environment friendly execution, millisecond-level fault detection, and fast response. The community management protocol for all-optical companies based mostly on centralized and distributed structure gives large connections, ultra-high reliability, and clever O&M.

4. Layered clever administration and management in addition to end-to-end orchestration. Using commonplace ACTN interfaces, the layered structure helps obtain cross-domain and cross-vendor E2E computerized orchestration and collaboration, offering open, quick, and layered optical community service provisioning and O&M, together with quick service provisioning, latency administration, and repair availability administration.

4. Integrated computing-network bearer networks

As IP networks evolve in direction of IPv6+/SRv6 alongside built-in computing-network structure, computing-aware networking can allow collaboration between cloud, networks, edge, units, and companies. Enabling applied sciences embody Application-aware Networking (APN), Service Function Chaining (SFC), and in-situ Flow Information Telemetry (iFIT).

The IPv6 extension header transmits software info to networks utilizing APN know-how. Using this info, the community can deploy companies and alter assets to fulfill SLA necessities of purposes. When a website is deployed on the fringe of a community (i.e., edge computing), APN know-how can join the community and purposes, adapt to the necessities of edge companies, and direct site visitors to the proper community path, bringing the benefits of edge companies into full play.

SFC – an ordered assortment of service capabilities – permits service chains to be created with various kinds of computing energy, in order that new sorts of companies may be rapidly provisioned. Service flows cross by specified value-added service units in a specified sequence to accumulate a number of value-added companies. In a computing community, SFC underpins intent-driven computing companies by connecting with totally different computing companies based mostly on buyer intention. Alongside SRv6 SIDs and associated companies, SFC helps construct a computing transaction platform. Computing energy offered by ecosystem companions is registered with the community as an SRv6 SID. Buyers should buy the computing energy they want, whereas the community connects computing energy companies by SFC and gives companies to patrons hassle-free.

Visualized computing paths and measurable efficiency have develop into key capabilities of built-in computing-network structure. iFIT helps correct on-demand packet-by-packet detection of efficiency indicators, resembling latency, packet loss, and jitter, for every service. Second-level telemetry information sampling shows the SLA of service flows in actual time. The hop-by-hop deployment mode permits ms-level fault restoration, guaranteeing the lossless switch of computing energy.

5. Customized companies with deterministic efficiency

Many vertical {industry} purposes, resembling industrial management in energy grids, manufacturing, IoV, and telemedicine, have exact necessities on community latency, reliability, timing accuracy, and information isolation. Increasing the adoption of 5G {industry} purposes requires differentiated, deterministic, and steady QoS for vertical {industry} clients.

Using the CUBE-Net 3.0 structure, China Unicom proposed offering custom-made networks with deterministic companies, which may be approached in the next two methods:

First, unified SLA parameters utilized industry-wide on deterministic companies. Key necessities of vertical {industry} purposes are categorised into totally different ranges to unify and standardize software state of affairs necessities, community indicators, and deterministic service indicators resembling latency, jitter, bandwidth, reliability, positioning, and clock precision. A visualized and user-friendly interface permits clients to observe companies towards SLAs in actual time and precisely sense service indicators.

Second, eradicating technical bottlenecks, particularly wi-fi air interface jitter, wide-area determinism, and the power to combine {industry} applied sciences. 3GPP has proposed URLLC know-how to supply low latency and excessive reliability over the air interface. A multi-service co-existence mechanism has additionally been launched. The feasibility and efficiency of each applied sciences are but to be verified. 

The capabilities of air interfaces should be enhanced. AI-enabled real-time management can offset the opposed results of getting numerous coexistent, advanced companies and mitigate the influence of site visitors burst on air interface efficiency. Currently, the applied sciences for delivering deterministic companies have many limitations. For instance, purposes like linked automobiles and distant management nonetheless wrestle to ship deterministic wide-area companies throughout areas. Delivering these companies locations excessive calls for on community gear and requires large-scale community upgrades, that means that value will develop into a limiting issue. Currently, 5G methods help latency-sensitive companies primarily by serving as a black field and integrating time-sensitive networking (TSN). Realizing native TSN capabilities and deeply integrating them with industrial networks is a key technical problem for custom-made 5G networks.

6. AI-enabled cloud-network mind

Advanced applied sciences like IT, information know-how, and AI may help construct safe, dependable capabilities like quick service provisioning throughout clouds or networks, domains, and distributors, built-in fault prognosis, useful resource visualization, computerized load balancing, and coordinated SLA monitoring and prediction. These will mix to type a CUBE-Net 3.0-based cloud-network mind.

End-to-end cloud-network scheduling and administration require useful resource abstraction and a scheduling engine.

Resource abstraction means abstracting clouds or networks in totally different eventualities and applied sciences into commonplace fundamental useful resource fashions based mostly on frequent information fashions and offering unified capabilities for buying, changing, storing, and managing assets.

Developing a service-independent computerized orchestration engine the place design and implementation are decoupled can (1) dynamically inject assets and repair fashions, (2) mechanically analyze service intent, obtain administration insurance policies and configuration objects mechanically transformed by B/O by scenario-based interfaces, intelligently management networks enabled by AI inference, and sense and reply to community standing modifications in actual time, and (3) help extremely dependable distributed transaction mechanisms, retries at failure breakpoints, and computerized or guide rollback based mostly on time snapshots. The orchestrator permits for visualization of assets at totally different layers, service evaluation and affiliation, and light-weight OSSs.

CUBE-Net 3.0 will present community digital twin administration capabilities with traceable standing, predictable targets, and straightforward adaptation to modifications. It may also provide digital help for community prognosis, prediction, decision-making, and sharing, and allow real-time interplay between logical and bodily objects, information privateness, and safety.

The real-time interplay between logical and bodily objects requires real-time sensing and dynamic modeling applied sciences that may assist exactly map logical and bodily networks. Software and AI applied sciences can allow interplay between bodily and logical networks, help with decision-making, and help steady enchancment. This will assist dynamically monitor and simulate the true standing, conduct, and guidelines of bodily networks, which can in flip help dynamic design and programming, fault simulation, and cutover simulation.

Digital twins create a brand new digital house exterior of closed-off bodily assets. This house will develop into more and more open and accessible, which can trigger extra safety and privateness points. Blockchains are decentralized, tough to tamper with, and supply non-repudiation, making them very best for information privateness and safety. CUBE-Net 3.0 will embody a blockchain-based trusted computing service system, and a extra environment friendly and dependable clever twin system enabled by AI.

AI is the technical basis for working and decision-making in autonomous networks. Going ahead, AI capabilities will probably be all over the place on the community ingredient layer. With environment friendly collaboration between on-premises and the cloud, the CUBE-NET 3.0 community structure permits AI fashions to take impact with only one click on and be replicated throughout your complete community, addressing the problem of large-scale replication of AI purposes that has lengthy plagued the {industry}.

The cloud has sturdy computing energy and may match all AI software eventualities throughout networks. As networks are deployed removed from finish customers, AI inference focuses on assembly time-insensitive companies (seconds or longer) and cross-domain/cross-vendor eventualities that require international collaboration, like service high quality forecasting over an extended cycle based mostly on efficiency indicators, IP and optical collaboration, and wi-fi and bearer community collaboration. Expert expertise and product information may be aggregated on the cloud to assist construct a robust AI coaching platform. Based on this platform, a community information repository may be developed utilizing information graphs and fed into on-premises, which incorporates the community gear layer and the administration layer, for clever community O&M.

On-premises are distributed throughout totally different geographical areas and are extremely time-sensitive. Due to restricted computing energy, they give attention to offering real-time AI inference capabilities. When a scarcity of samples of on-premises can meet the precision necessities of coaching fashions, the fashions may be retrained on cloud and the outcomes fed into the on-premises inference framework. To tackle the growing older of on-premises fashions, an everyday replace and evaluation mechanism is used to pick out optimum fashions.

7. Data-driven clever operations

Data-driven clever community operations comprise three elements: information convergence and the opening up of capabilities, information intelligence and constructing the driving force layer, and modern information purposes.

The first half is information convergence and the opening up of capabilities. All-domain convergence permits information supply identification and administration based mostly on standardized and ID-based information and helps the institution of information evaluation chains and information lineage tracing. Scenario-based information aggregation and public information fashions may help construct information interfaces which can be scenario-based, standardized, automated, and that translate information into info. The layered loop gives intent-based interfaces externally, so the higher layer can have decrease community experience necessities. Unified APIs and information make it simpler for the higher layer to accumulate community information, serving to construct a layered information governance structure that helps information assortment, transmission, and the evaluation of various ranges of time sensitivity at totally different scales and totally different ranges of precision.

With an information safety administration framework that covers your complete information lifecycle, information administration focuses on the highest-risk information transfers. A safety administration coverage created based mostly on a evaluation of current information property and information dangers can rapidly fulfill the custom-made information safety wants of companies and households.

The second half is information intelligence and constructing the driving force layer. General algorithm fashions, information simulation, and a community operation information repository are used to construct a data-driven center layer for the analysis and growth of upper-layer purposes.

The third half is modern information purposes. Using converged information and open capabilities, community AI algorithms, and community operation information repositories, we are able to construct a sequence of clever information purposes resembling dynamic community monitoring and prediction and community configuration, upkeep, and optimization. These are used to trace and sense service objects in actual time whereas dynamically and intelligently adjusting assets to fulfill numerous buyer wants.

8. Intrinsic cybersecurity

The CUBE-Net 3.0 intrinsic community safety module makes use of the next 5 applied sciences:

Channel-associated trusted id know-how optimizes communication protocols and community gear, and embeds trusted identifiers and password credentials into packet headers. Network gear can confirm the authenticity and legitimacy of requests by id verification, stopping fraud and constructing fine-grained entry authentication and sourcing capabilities.

Intrinsic NE safety know-how based mostly on trusted boot and irregular conduct detection introduces chip-level trusted computing know-how into NEs, which helps construct a trusted, safe software program and {hardware} working surroundings based mostly on the underside layer of NEs. This permits for verification in any respect layers, from {hardware} platforms and working methods to purposes, guaranteeing the confidentiality and integrity of methods as an entire.

AI-enabled dynamic planning of safety insurance policies strengthens capabilities like self-learning and the modeling of site visitors and repair traits, threat prediction, and safety coverage orchestration based mostly on characteristic fashions. This perform additionally permits safety coverage battle detection and computerized optimization.

Blockchain-enabled safety administration of digital community assets helps construct a trusted system for digital community assets resembling IP addresses, domains, and AS numbers. A distributed accounting and consensus mechanism ensures the authenticity of useful resource possession and mappings and prevents safety points resembling IP tampering, route hijacking, and area spoofing.

Software-defined and built-in safety functionality orchestration gives safety capabilities in the type of swimming pools and microservices, making built-in orchestration and the opening up of capabilities doable. This will permit customers to flexibly outline safety insurance policies, invoke safety assets on demand, and agilely deploy and roll out safety capabilities. Intelligent coverage planning permits safety on any cloud and any community.

China Unicom,Network Architecture

Enabling {industry} digitalization with an open ecoystem

China Unicom will use CUBE-Net 3.0 to create an ecosystem with capabilities opened up on demand and simply built-in, construct a unified platform for opening up community capabilities, implement a technique of community service as a platform, present standardized community functionality APIs, open up extra community service parts in the type of purposes, and make community as a service (NaaS) a actuality.

It may also work intently with vertical industries, {industry} software builders, and {industry} gadget suppliers to combine purposes, computing, networks, and units, and supply clever, converged companies for vertical industries, thus build up its strengths in connectivity, computing, and intelligence, enabling {industry} digitalization.

With CUBE-Net 3.0, China Unicom may also work to construct a controllable open supply know-how ecosystem. We will proceed to pursue open innovation, make full use of built-in purposes and consolidated assets, broaden the open supply ecosystem, and work with {industry} companions to construct a technical alliance for the open supply {industry}. To develop into self-sufficient in fundamental software program and {hardware} and core applied sciences, China Unicom will intensify efforts to pursue technical partnerships with home leaders. 

Together, we’ll construct a full-stack open-source know-how ecosystem that covers chips, units, networks, working methods, and safety, over which we can have full autonomy.

https://www.huawei.com/us/technology-insights/publications/huawei-tech/202202/8-innovations-china-unicom-2025-network-architecture

Recommended For You

Leave a Reply