您当前的位置是:  首页 > technology > Corporate communications > Abstracts >
当前位置:首页 > technology > Corporate communications > Abstracts > Huawei Fellow Zhu Guangping talks about autonomous networks and services

Huawei Fellow Zhu Guangping talks about autonomous networks and services

2018-05-16 09:52:11 Author: Zhu Guangping Source:CTI ForumComment:0  Click:


During the Global Analyst Conference in 2018, Huawei announced the SoftCOM AI solution architecture and introduced AI technology based on a fully cloudized network. It strives to create an "always-on fault" autonomous network and set off a new round of network changes.
Cross-border competition calls for system architecture innovation
We are entering a new era of cross-border competition. Each industry faces structural challenges, especially for the telecommunications industry.
First, from the revenue structure, the operator's business is facing challenges from the IT industry. Previously, telecommunications business was divided into three levels: terminal, network and IT infrastructure and upper application. With the dramatic increase in network access rates, the IT industry has changed from selling products to selling services, and backbone networks and IT infrastructure have gradually become forms of cloud services. Operators can compete with AWS and other cloud service giants to compete for the trillion-dollar cloud market if they can do a good job of cloud services. Otherwise, they will lose a lot of traditional telecommunication services, especially the dedicated services between data centers, such as Ali. Cloud is building a cloud backbone that connects the world.
Second, the efficiency and cost of operators also face structural challenges. Today, OPEX spending on telecom equipment maintenance is about three times that of CAPEX, putting a heavy burden on operators. In addition, the telecommunication network is more and more complicated, exceeds a person's professional knowledge and ability, causes 70% of major network breakdowns to cause the artificial factor to cause. As the chief wireless architect of the Canadian operator TELUS stated: "Machine manufacturing is becoming automated, and the telecommunications industry is still in the stage of handicrafts."
To solve the challenges faced by the telecom industry, product innovation alone is not enough. It requires the innovation of the entire system architecture and the innovation of business models to enhance the competitiveness of operators and solve structural problems. What is system architecture innovation? Taking cloud computing as an example, it is not an innovation of a server or a storage product. It is a system-level innovation that uses a new distributed system to increase resource utilization efficiency. Product innovation, system architecture innovation, and business model innovation support each other and promote each other.
In order to meet the needs of customers in the new era, Huawei's innovation system is designed in accordance with the above three dimensions. At the product level, the guiding ideology of Huawei's design of network equipment is the “Olympic spirit”, which means high capacity and low latency. All product innovations are centered around this goal. In the field of system architecture innovation, Huawei aims to build an agile automated and intelligent network and implement the network's "autopilot mode." There are two goals in business model innovation. The first is to become one of the five clouds in the world by providing cloud services. The second is to build an online intelligent service model in the Internet era.
SoftCOM AI brings new value
Recalling the development path of Huawei's network architecture, we proposed the Single strategy in the ALL IP stage. After the rise of cloud computing, we entered ALL Cloud stage in 2012. We propose SoftCOM to implement a data center-centric network. In recent years, with the development of artificial intelligence technology, we have proposed to be fully intelligent (ALL Intelligence), introducing artificial intelligence into telecommunication networks. SoftCOM AI was thus born with the purpose of realizing autonomous networks at the network architecture level and services at the business model level. 2.0.
The introduction of an artificial intelligence autonomous network aims to build "Industry 4.0" in the network field and realize "autopilot" of the network. Industry 4.0 has three features, namely, agile devices, intelligent controls, and intelligent analysis systems to automate production. This is also true for the telecommunications industry. In the telecommunication network, the lower layer is the network equipment and the upper layer is the control layer. In the aspect of the control and operation and maintenance of the entire network, the artificial intelligence technology is introduced from end to end to construct the segmented autonomy function. Each section of autonomy is realized through the upper layer operation system. End-to-end autonomy, which in turn enables network-wide autonomy. The biggest change brought about by the autonomous network is that the maintenance personnel are not an automated system in the entire business process. We call it the “automated network driving mode” to realize the self-optimization, self-healing, and automation of the entire network.
The goal of Service 2.0 is to create an "Industrial Internet" in the network domain and provide online digital "smart services." This service concept is extended to the telecommunications industry. The future network is fully automated at the operator side. Huawei provides fully automated online services based on artificial intelligence in the background. This service is based on a continuous iterative model and builds a model based on industry practice. That is, the service is always in the beta phase and is continuously updated and improved.
The introduction of AI into telecommunication networks brings new value to "predictability." The management and control center of the telecommunication network is based on the southbound interface of the device and data collection. Through certain strategies and rules, the entire network can be managed and scheduled. The basis for its implementation is mainly three conditions, including network accessibility, SLA requirements and resource efficiency, which are the basis for network automation. However, with the increasing complexity of networks, these are far from enough. It is necessary to introduce algorithm-based network management and control, online AI reasoning, and data analysis in the network to achieve traffic prediction, quality prediction, and failure prediction. Predictability is the core value of AI. It is based on unknown conditions to schedule the network, avoid faults before failure occurs, optimize quality before quality deterioration, and adjust traffic before network congestion, so as to achieve automatic, self-optimal, self-healing, and autonomous failure. The autonomous driving network has structurally improved operation and maintenance and operational efficiency.
Improve user experience and achieve three multipliers
To realize the automatic driving of the network, it will inevitably be a long-term process, and it cannot be accomplished overnight. With reference to the five development steps of auto-piloting, we also divided the autopilot network into five phases. The first phase is where AI can indicate “what happened” and the second phase needs to determine “why it happens”. Stages need to predict what will happen. Follow-up will need to make decisions manually and take appropriate measures. In the fourth stage, AI can already determine what measures need to be taken, and then proceed manually; the last stage is Fully realize network self-control and automatic repair, so that the network has self-healing capabilities.
The realization of autonomous network and service 2.0 will bring to the end user a minute-grade ROADS experience, always optimal network connectivity and permanent network availability; the value to operators is to achieve operational and maintenance efficiency, resource efficiency and energy efficiency. The multiplier of efficiency.
In the operation and maintenance field, the operation and maintenance level is divided into three development stages. The first stage is called R2F (Run-to-Failure). The network suddenly malfunctions during operation, and the operation and maintenance personnel immediately rush to handle it. This is the lowest. Level; the second phase is PvM (Preventive Maintenance), which is a routine patrol. Each device is inspected to prevent failures, but this is inefficient; the third phase is PdM (Predictable Maintenance). ), we call it predictive maintenance, that is, we can predict the probability of a certain device failure in the future, and then carry out targeted maintenance. Through PdM, we hope to reduce the alarm compression and fault location of telecommunication networks by 90%, achieve 90% failure and degradation prediction of key devices, and further achieve network self-healing. In addition, more than 70% of network faults originate from passive devices, such as optical fiber bending aging, loose interfaces, etc. In this process, the signal will change. By introducing AI to learn the characteristics of these changes, it is possible to advance. It is predicted that active passive faults will be solved.
In terms of network resources, the current feature is that the network is well-established, traffic will flow, and the use of resources may not be reasonable. If you think in reverse and schedule the network based on the flow direction, resource utilization will increase significantly. Today's networks do not have such capabilities. Only by introducing artificial intelligence and building a traffic prediction model can we achieve accurate traffic prediction and the most reasonable network topology. We can use the traffic instead of the physical connection to determine the path of the network, and ultimately increase the network significantly. Resource efficiency.
Regarding energy efficiency, there is a saying called “bits determine watts”, that is, the amount of network traffic determines how much energy is consumed. In the engine room or site, each system has dozens of parameters, and the heat and environment and service load models are generated through AI training to achieve the best energy efficiency for sunshine, temperature, and auxiliary facilities such as oil, solar and batteries; at the equipment level, According to the business load dynamic energy delivery, when there is no traffic, the use of time slot off, RF deep sleep, carrier frequency off, etc. to reduce power consumption, while achieving dynamic energy management of data center objects such as server components; the third is the network The system constructs an accurate business load forecasting model so that the entire network traffic is optimized to achieve the best energy efficiency.
The target architecture of the autonomous network is SoftCOM+AI for Huawei. The specific approach is to plan, deploy, operate, maintain, and manage the underlying devices and cloud infrastructure, the network management and control at the middle tier, and the upper-level system. In the end-to-end process of optimization and management, artificial intelligence technology is introduced in each link to enable the network to achieve optimal system. At the same time, Huawei also built an operator-oriented AI training platform to train the AI ​​model on the state data access platform of network equipment, and continuously update and optimize the model so that the degree of automation of the network system is continuously improved.
Taking optical networking as an example, let's take a look at how AI can enable the whole process of business development. The first is the data base, that is, what kind of data needs to be obtained. Specifically, the optical network includes optical fiber data, optical signal data, and routing data. The next step is the enabling technology, namely, artificial intelligence algorithms, including data cleaning and information. Integration, machine learning modeling, deep learning, etc., these algorithms are not related to optical networks; in order to realize automatic driving of optical networks, a large number of models, such as fiber optic models and filter models, need to be built; finally, the business application scenarios, including the start Automatic inspection of optical fibers, service provisioning, network optimization, fault location, and automatic resource scheduling, etc., to find the best method through the model, and then achieve rapid release, minimal operation and maintenance and intelligent operations, intelligent to improve the efficiency of network scheduling, zero wait, zero contact, zero Experience, so that people do not feel the existence of the network.
The future will be an era of intelligence. The intelligence of carrier networks cannot be achieved overnight, but it is a long-term practice. SoftCOM AI is Huawei's All Intelligence strategy in the field of telecommunications. The core AI capability is based on Huawei's long-term and resolute strategic investment in All Intelligence. It is integrated with the telecommunication field scenario and is designed to help operators create new opportunities. Faulty autonomous networks realize digital and intelligent transformation as soon as possible.
[Disclaimer] This article only represents the author's own opinion and has nothing to do with the CTI Forum. The CTI Forum maintains its neutrality in the presentation of statements and opinions, and does not provide any express or implied warranty for the accuracy, reliability or completeness of the contents contained therein. Readers are for reference only, and please bear full responsibility for yourself.

Topics