AI-Driven Assurance Is Critical for Ensuring 5G Service Quality

November 24, 2020

Introduction

5G is ushering in a more wireless world in which we are always connected, services are constantly available on-demand where we can do more on the go, and tasks that in the past would take minutes now take seconds. However, the shift to a 5G network is a seismic transformation of the network architecture and its management. Adding to this is the significant increase in data traffic and millions of more connected devices. In this environment, operators will have substantial challenges in monitoring and managing their networks if they continue relying on traditional manual processes. Sifting through the vast amounts of data to understand the customer experience and optimize network performance, will require a new approach. In the end, the success of 5G depends on whether operators can manage these networks more efficiently while delivering the promise of a more personalized customer experience.

Looking for a needle in a field of haystacks

Mobile networks used to be more straightforward. The only service on 2G was voice, designed and packaged for the telecom operator’s subscriber. To ensure service quality, network engineers needed to monitor a handful of different Key Performance Indicators (KPIs) to provide the service was running correctly. As more and more mobile technology generations have rolled out, more and more complexity has been added. Now, hundreds of different services are running over the network, voice services (typically IP-based VoLTE) and data services that cover a wide range of use cases from video streaming to social media platforms. Monitoring all this data and pinpointing customer-affecting network degradations will be like looking for a needle in a field of haystacks. With so many different services, devices, network traffic, and combinations, traditional service assurance solutions with manual processes will fail in 5G.

70% of operator time is devoted to the discovery and root cause of network issues.

Analysys Mason

With operators already spending significant amounts of time isolating service degradations and troubleshooting the network, how are they expected to manage the transition to 5G and the clear jump in the data load while managing the network more dynamically?

It is apparent that for 5G, the traditional telecom network operations and management model will not meet the increasing requirements needed to ensure a smooth transition to 5G. The 5G era requires an intelligent, more automated network. This will allow operators to manage their networks more effectively and deal with the increasing complexity and enable innovative business models, like network slicing and the dynamic management of networks, to provide a more personalized customer experience. Introducing intelligence into the network through Artificial Intelligence (AI) is the key that will unlock the ability to offer quality, dynamic services at scale and deliver on the promise of 5G.

AI is the key

With the development of fast, low-cost computing and cloud technologies, machines can analyze massive amounts of data generated by networks every day and embed it into next-generation service assurance solutions.

AI has natural advantages over humans in analyzing massive amounts of data and finding patterns and relationships in the data. Machines:

  • Handle repetitive assignments
  • Process complicated, multi-dimension tasks
  • Process and correlate information from many sources
  • Do not require manual adaptation
  • Accumulate experience over time
  • Work non-stop 24/7/365

This will free engineers up to spend more time on the critical task of optimizing the network performance and solving network degradations, rather than wasting time looking for needles, machines and humans working together to ensure superior customer experience and operational excellence.

AI insights for 5G

AI will be essential for network operations and managing the customer experience in 5G. Furthermore, for more advanced 5G services like dynamic network slicing creating rules and policies that proactively prevent and resolve issues will be vital. All this is part and parcel of the operators’ goal to enabling an open/closed-loop approach to network management. Automated insights provided by AI will feed into the operators’ orchestration to allow this approach, which saves on OPEX and ensures the network quality automatically, which will be required to deliver high quality, personalized services in the 5G era. So, how do operators best acquire these AI-driven insights into their network?

Embedded into AI-driven assurance

By deploying cloud-native service assurance solutions with built-in AI/ML, the operator can utilize the data already collected through the solutions’ containerized probes. AI/ML is then applied as the data is collected rather than deploying an additional solution for AI purposes.

Having service assurance with built-in AI offers several benefits to operators.

Data already collected for assurance is used:

  • AI is applied to all data collected and not a subset
  • Saves unnecessary expenses for an additional solution (such as storage costs)
  • Saves time massaging the data for external processing
  • Runs on any data set (for example, first throughput and instantly change to the release cause)

However, these next-generation assurance solutions need to come with modular architectures that are machine-learning-friendly. New and updated ML models can be seamlessly updated and integrated into the solution. Best-in-breed algorithms and ML models such as Prophet for forecasting time series data (built and open-sourced by Facebook) should be used and offer the ability to choose from various algorithms and ML models so that the right ones can be used for the specific use case. Here are some of the use cases in which operators can use AI-driven assurance to assure their networks automatically.

Optimize network performance

Anomaly or outlier detection is a method of searching for data that does not match an expected behavior or a pattern in a given data set. An example of KPI-based anomaly detection is for release cause. An advanced algorithm is applied to identify anomalies in the release cause count between all network elements and the associated severity over time. It is thus removing ‘outliers’ from the release cause baseline.

The baseline outlier removal facilitates an accurate baseline prediction, improves the detection of network anomalies, and removes false positives. Also, in calculating a baseline, and a confidence area, it is possible to see points in time beyond the confidence area based on past data. These exceptions can be translated into additional alerts. By utilizing these AI-driven insights, engineers remain focused on handling critical customer-affecting issues rather than handling “alerts” that have no real effect on subscribers.

Smartly plan network capacity

After generating baselines in different network performance and service quality areas, AI-driven assurance can use this information for more use cases. The most common is predictive analytics. As the solution already knows how to generate the right forecast in the short-term (for example, the release cause KPI), it’s also possible to create a long-term forecast for various performance and quality indicators and use it to plan future network capacity smartly.

By building these processes at the early stages of 5G, engineers will learn to trust the insights that AI can provide and integrate them more into network management, allowing more data-driven decision-making that will be critical for 5G network management operations. Furthermore, by incorporating more sophisticated AI algorithms, which consider seasonal seasonality, it will also be possible to create forecasts for an extended period of time. For example, one month or a whole quarter.

Improve the customer experience

With more and more network traffic being encrypted, operators face the challenge of understanding the customer’s Quality of Experience (QoE). By utilizing AI, ML, and heuristic modeling, AI-driven assurance can shine a light into the darkness and provide insights into the customer experience for multiple use cases, such as video streaming. By taking measurements from both subscribers and the network and applying machine learning algorithms to match those two data points, AI-driven assurance can generate quality of experience metrics, such as:

  • MOS-B video quality
  • Start delay
  • Effective and network video payload bytes and packets
  • Rebuffering indication
  • Effective video throughput

These metrics could be evaluated for all customers in the network, not only for specific users that were analyzed. So, by training network performance metrics, with real-life customer measurements using machine learning, a complete view of customer experience can be generated for each specific service. Different regions are affected in different ways, and traffic usage varies. So, these metrics can be generated for each distinct network.

In conclusion, to ensure 5G network quality, operators will need to deploy cloud-native assurance solutions with built-in AI/ML. This will provide the most efficient way to embed artificial intelligence into the network to aid network engineers in managing their networks and separating the network “wheat” from the chaff. This will free engineers up to spend more time on the critical task of optimizing the network performance and solving network degradations to ensure superior customer experience and operational excellence in 5G.

This blog was published as part of Light Reading’s Partner Perspective series.

Share this article
Share on twitter
Share on linkedin
Skip to content