Premium Broadband With Intelligent Home Broadband O&M, Reshaping Network Experience
With the popularization of new home broadband services such as 4K, VR and online gaming, users' requirements for home broadband services are broadening from solely network speed to multiple dimensions, including ultra-broadband network, ultra-clear and smooth video, ubiquitous WiFi coverage, low-latency and prompt troubleshooting.
Traditional complaint-driven O&M is passive and inefficient. It relies on door-to-door maintenance and manual experience, severely hampering O&M efficiency and user satisfaction. In addition, operators cannot view network resources in real time, are unaware of home broadband network experience and lack effective user retention methods. O&M costs (opex) increase with network scale, sometimes amounting to two thirds of operators' service revenues. This prevents network potential realization and causes high user churn rates.
Introducing artificial intelligence (AI) and big data analysis to slash opex
Facing these challenges, AI and big data analysis are key technologies to break the curse of OPEX growth proportional to network scale. In the industry field, AI has been widely used for facial recognition, license plate recognition, translation and healthcare. In the communications field, leading telecom operators in the US, Japan, the UK and China have begun exploring opportunities of AI and big data in fault diagnosis and prediction.
AI and big data analysis in the telecommunications field brings the following changes to operators' operating modes:
- Realize overall quality analysis and optimization by leveraging high-precision data collection and big data analysis instead of the traditional fragmented approach.
- Construct digital twins of existing physical networks based on collected 360-degree massive network data to simulate networks without affecting services, evaluate the impact of new services on networks in advance, and expand network capacity to reduce risks. Operations are centered on users instead of equipment, and are directly connected to customers' business intentions.
- Build a massive fault diagnosis expertise library based on O&M experience of the entire industry, use machine learning and AI algorithms to realize intelligent and precise fault locating and prevention, replacing inefficient traditional troubleshooting, which relies heavily on individual experience.
Based on these considerations, Huawei has launched the industry's first premium broadband Intelligent Home Broadband O&M solution based on the Network Cloud Engine (NCE) platform. The solution features a three-layer distributed edge computing architecture, Telemetry fast data collection technology, hardware data collection chips and industry-leading AI algorithms (PON optical path characteristics algorithm, WiFi channel optimization algorithm and optical path failure prediction algorithm). In addition, to help operators build future-oriented network O&M, Huawei has proposed many innovative methods such as WiFi quality optimization, intelligent ODN fault diagnosis, proactive rectification of weak light problems, network bottleneck prediction for proactive capacity expansion and identification of users with poor experience.
Visualizing and optimizing home network quality to improve user experience
WiFi eliminates the need for indoor cabling and enables new services such as multi-screen and cloud VR, with the side effect of more factors affecting broadband experience. Factors include WiFi signal interference and insufficient coverage, which can vary from user to user. As operators cannot visualize network resources and connection status, more than 30% of faults require home visits for troubleshooting. Among these, 25% are sporadic problems such as slow network speed and sporadic video freezing. These faults are difficult to recur, resulting in repeated user complaints and home visits. Each trouble ticket can take as long as two days to handle.
To address such sporadic problems, Huawei has launched the NCE platform to enable operators to visualize home network topologies and real-time connection status. Within seconds, the platform collects from ONTs more than 20 characteristic parameters such as channel status, negotiation rate, neighbor information, interference duty ratio and background noise. The collected data is analyzed based on Huawei's global million-level fault library, 250 home segment fault models, and TBs of seven-day network data. Network status can be detected, and historical quality issues can be played back with one click. This allows most home network faults, such as WiFi interference, insufficient coverage, connection failures, configuration errors and device faults, to be identified in minutes and remotely rectified by customer service agents. For long-term WiFi signal interference problems, historical big data analysis can be performed to train models based on typical faults, enabling the system to learn and predict optimal channels and perform channel optimization periodically. This ensures user experience while reducing manual intervention and the fault rate.
Intelligent ODN optical path diagnosis, accurately identifying faults and reducing opex
Currently, operators use two fault demarcation methods on PON optical paths. The first method is to locate faults based on alarms and connection relationships between users and optical splitters in the inventory system. However, because resource data can be inaccurate, optical path faults, especially non-disconnection faults in distribution fibers, are difficult to locate and have become O&M pain points for operators. For such faults, trouble tickets are often dispatched to incorrect handlers, and troubleshooting can be slow. The second method is to locate faults segment by segment using a hand-held optical time domain reflectometer (OTDR). This method increases O&M costs, and for users who do not report faults but experience weak light, onsite rectification is required, which can take months in multiple batches. Because operators cannot prioritize rectification, rectification is inefficient.
For ODN optical path faults, the Huawei premium broadband Intelligent Home Broadband O&M solution collects 20 types of optical path characteristic parameters, such as optical power, bit error rate and optical distance, from ONTs and OLTs in seconds. Then, the solution implements big data analysis on these parameters; extracts optical modules characteristics such as steady state, jitter and trend; and clusters ONTs with the same characteristics to learn the ODN topology of the entire network. Based on the AI PON optical path fault characteristics decision tree, Huawei's million-level fault library and more than 150 access fault models, the solution identifies faults and locates root causes with an accuracy up to 80% by continuous comparison and training. More than 20 types of faults, such as fiber connector loose or contamination, overbending and disconnection, can be identified and clearly marked in the ODN topology.
In this way, trouble tickets can be accurately dispatched, ensuring troubleshooting efficiency. In addition, the solution can be interconnected with operators' work order systems to allow trouble tickets to be automatically created based on preset thresholds. For users who experience weak light, optical path information can be collected at the receive and transmit ends, and AI algorithms can be used to identify root causes, demarcate faults and associate faults with user service experience. This helps prioritize rectification, greatly improving rectification efficiency and reducing potential complaints.
Exploring network data to identify users with poor experience, evaluate network capacity expansion requirements and support efficient operations
Currently, network data is static and fragmented, and operators lack intelligent data analysis tools to support efficient network operations. Users with poor experience cannot be identified to support up-selling. Moreover, network bottlenecks cannot be predicted and evaluated to support network capacity expansion.
With the help of the Huawei NCE platform, operators can directly observe home network topologies and real-time connection status; collect indicators such as interruption duration, number of interruptions, packet loss rate, Wi-Fi coverage, and interference; and determine the poor experience threshold by using bias and outlier algorithms. In this way, operators can identify poor experience phenomena using the voting algorithm, and conduct user retention to lower the churn rate, especially for VIP customers.
What's more, the NCE platform can group users with different service features (such as bandwidth, service type and usage frequency), sample the average peak bandwidth of each group over time, and establish prediction models to forecast traffic. In this way, the network health of operators can be evaluated, and network congestion can be evaluated for service upgrades such as broadband acceleration and video bitrate improvement. This evaluation can be correlated with the user development trend to support precision capacity expansion. The NCE platform supports multi-session parallel processing and 100 Gbps data throughput, and can process tens of millions of NE data records. In combination with AI and cloud-pipe-device data telemetry, data can be collected quickly, making data analysis more reliable than traditional fragmented network analysis. The analysis results better meet personalized analysis and application requirements of various network operation departments.
In the premium broadband innovation program jointly conducted with selected area of 400,000 users of an operator, Huawei helped the operator reduce annual troubleshooting time by 50,000 hours, reduce annual number of home visits by more than 20,000, improve O&M efficiency by 50%, and reduce the user churn rate by 30%. It also reduces estimated OPEX by US$1 million annually.
The Huawei premium broadband Predictive Intelligent Home Broadband O&M solution can play back historical broadband events in homes, automatically identify root causes of PON optical path problems, intelligently optimize WiFi channels, identify users with poor experience, evaluate network health, and predict network congestion. These features enable operators to shift from device-centric to user-centric business models, strengthening differentiated competitiveness, preemptively identifying and resolving network problems and providing users with new and better broadband experience.