PARTNER FEATURE: In the run up to the GSMA’s inaugural Thrive Africa event later this month, ZTE has released its Autonomous Evolving Network white paper. The white paper comprehensively introduces the architecture, evolution objectives and paths towards true automation, including 10 valuable scenarios showcasing the autonomous evolving network.
CommsMEA caught up with Jason Tu, principal scientist of NFV/SDN solutions at ZTE, to hear his insights into the evolution of intelligent networks.
Jason TU, Principle Scientist of NFV/SDN Products, ZTE Corporation
Can you tell us a little about ZTE’s understanding of the progress and requirements of network intelligence?
Jason Tu: It is a long story. About 10 years ago, ZTE started its research and development into 5G. We found that the bandwidth of the average 5G user is at least 10 times that of 4G, and the number of IoT-oriented 5G terminals will also be several times that of 4G. The network will be more complicated and difficult to operate. Obviously based on the traditional O&M mode, operators have to hire more engineers with higher skills.
Firstly, ZTE started to build an “Elastic network” based on NFV/SDN and cloud native technology, which can simplify the network and deal with network capacity and performance. Then in 2018, based on the Elastic network, ZTE introduced the big data analysis and preliminary AI technology to release the uSmartNet Network Intelligent Solution, which mainly achieved autonomous network O&M and greatly reduced O&M manpower and complexity.
In 2020, ZTE combined extensive practice experience in global telecommunication networks with advanced AI engines and capabilities such as deep learning and RCA (Root Cause Analysis) to implement automatic network optimisation, self-organisation and evolution. AI and Big Data-based network O&M knowledge Library will greatly simplify the complexity of network O&M and effectively avoid human network failures. Our Autonomous Evolving Network white paper covers operators’ major key problems in the stages of telecommunication network planning, deployment, optimisation and O&M. It focuses on ten valuable scenarios, including Base Station Energy Saving, AI-Based M-MIMO, Base Station Troubleshooting, Smart PON, Smart CDN, Smart Bearer Network, Smart MEC(Multi-access Edge Computing), AI-Based NFV O&M and Smart Slicing and AIVO (AI Insight Value Operation).
What role will AI play in network intelligence solutions and what is ZTE’s capability in terms of its AI platform?
Jason Tu: AI technology is introduced to telecommunication networks in order to analyse the massive amounts of data generated from network operations and to achieve more efficient and automatic network O&M, thereby reducing the manpower required. At the same time, when the new generation AI engine and algorithm are introduced, the network can accumulate a large number of policy libraries to deal with unexpected situation, and can even perform self- optimisation and Autonomous Evolving. To introduce AI into the communication network, hardware acceleration capability, AI learning framework, AI algorithm, AI application, and other capabilities should be supported.
ZTE has extensive experience in the following areas:
Hardware: ZTE supports GPU, CPU, FPGA or HPC cluster.
AI framework: ZTE supports mainstream deep learning framework and machine learning framework, and research and develop key technologies such as parallel computing acceleration and inference prune.
AI algorithm components: ZTE has explored and optimized over 100 network-related algorithms, covering algorithm models such as radio network, bearer and core network.
AI application components: ZTE provides specific 5G-oriented intelligent applications, and flexibly supports a variety of application scenarios, such as intelligent prediction, RF fingerprint, smart slicing and RCA (root cause analysis).
Tell us a little about the changes in network capabilities and the work of O&M personnel after the introduction of these solutions?
Jason Tu: The introduction of the Autonomous Evolving Network Solution can greatly improve network capabilities and reduce the workload of O&M personnel. There are several examples.
A) Energy saving of wireless base stations
Traditionally, O&M personnel use relatively fixed policies to enable and optimise the energy saving feature based on human experience and data analysis. After the AI energy saving solution is introduced, the platform automatically analyses the network parameters and KPIs, and automatically completes the initial parameter configuration, threshold adjustment and closed-loop optimisation. In addition, it can generate differentiated policies for different sites, so as to match the energy saving requirements of the whole network.
At the end of 2019, ZTE and the Shandong branch of China Unicom jointly completed the comparative verification of energy saving of 10,000 cells in the existing network environment. It took one O&M engineer a week to build the AI platform and deliver tasks. Later, the AI automatically adjusted and optimised the energy saving plan, while the O&M personnel carried out daily monitoring for about one week to achieve the best network performance and energy saving effect. In general, it takes one person two weeks to optimise the network. By contrast, it would take two O&M engineers three months to perform batch data analysis and optimisation to achieve the same effect as that of AI optimisation. Thus, the workload is reduced greatly after the introduction of AI. Besides, according to a calculation, more than 21,000 kilowatts of electricity can be saved per week for these 10,000 cells, equivalent to 20.6 tons of carbon emissions per week. It is estimated that more than 10 per cent of energy can be saved for the whole network.
Since mid-2019, base station AI Energy Saving has been put into commercial use in more than 10 operation networks in over 100,000 cells in countries such as China, Malaysia, South Africa, Italy and Indonesia.
B) Bearer network configuration audit.
Telecom operators are facing the situation of 5G network construction and upgrade. The basic configuration of the network and introduction of new functions will bring a huge workload to the O&M personnel, which may result in missing configuration of parameters, incorrect configuration and configuration conflict. ZTE has developed the intelligent configuration check function based on NLP (Natural Language Processing) and knowledge graph. The function can check the configuration and find problems before they take effect. In a recent exercise with the Guangdong branch of China Unicom, the automatic configuration check shortened the traditional process from 90 man-days to 7 man-days, and reduced the incidence of failures caused by abnormal configuration by 85 per cent, thereby significantly raising the operation and maintenance efficiency. This feature has won the Best Network Intelligence Award at the Broadband Awards 2019.
C) ZTE AIVO (AI Insight Value Operation) digital operation solution
AIVO solution constructs an integrated system of planning, maintenance, optimisation and business operation. It uses a hierarchical AI engine to integrate NOC and SOC with big data and AI technologies. In this way, automatic closed loop for full-service can be realised. Through data association, AIVO realises the management of the NE layer, network management and control layer, and service operation layer, while simultaneously providing the integrated O&M capability of “unified view, full-domain perception, intelligent closed-loop and proactive O&M.”
In one VoLTE project, the AIVO solution provides O&M support for more than 100 million VoLTE subscribers and promoted the increase of millions of subscribers per month. At present, ZTE AIVO digital operation solution has been deployed and applied in over 30 networks in China and other countries.
What about ZTE’s deployment of Autonomous Evolving Network in Africa? What benefits have you brought to African customers?
Jason Tu: ZTE has delivered a whole range of AI solutions to operators in Africa, including network performance analysis, traffic prediction, precise planning and energy saving applications. The value and effect of these solutions has been verified by operators like MTN South Africa, Airtel Nigeria, Ooredoo in Algeria.
For example, ZTE has developed the AI-based intelligent power saving solution for 4G networks in MTN South Africa. The solution uses AI algorithms to implement accurate traffic predictions, thus improving the actual effective time of energy saving of 80 per cent. Compared with the traditional energy saving solution, the AI-based intelligent power saving solution saves an additional 4.08 per cent of total energy consumption while maintaining the same network performance.
You can download ZTE’s white paper, Autonomous Evolving Network, by clicking here.