Analytics to build a resilient 5G network

A 5G network is a software-defined network.

This switch will allow MNOs to compete in the 'big data' analytics space. But they will need new machine-learning-based tools to do so.

Thales stands ready to aid MNOs in predicting future network events and customer behaviors, integrating crucial 5G analytics into their strategies.

The adoption of 4G has fuelled a huge consumer appetite for data-driven services. Today, the average smartphone user consumes 10 gigabytes of data per month.  

As a result, global mobile data traffic (excluding fixed wireless) exceeded 49 exabytes (that's 49 billion gigabytes) per month at the end of 2020. 

This is just the entrée.

According to Ericsson, the monthly total will hit 237 exabytes in 2026.

However, the traffic generated by the Internet of Things will dwarf even this total. A study by IDC estimates that 41.6 billion connected IoT devices will be connected by 2025. Collectively, they will generate an eye-watering 79.4 zettabytes (ZB) of data. That's 75 trillion gigabytes. 

What's fuelling these extraordinary projections? 

Obviously 5G. 

By leveraging the high speed and low latency of 5G and employing 5G analytics, MNOs can gain real-time insights, enhance service quality, and drive strategic decisions in network management and customer engagement.

More data = more opportunities to predict future events 

The new standalone 5G data infrastructure represents a break from the existing cellular world. A 5G network is a software-defined network capable of exponential speed and capacity improvements. 

It can support up to one million devices per square kilometer, and transmit data at 10 gigabits per second. 

Self-evidently, the primary mission of the world's MNOs in a 5G data world is to connect tens of billions of devices reliably. They must then protect the identity of these endpoints – and secure all those zettabytes of 5G data. 

However, there is something else they can do with all that traffic. They can analyze it – using new machine learning tools – and use the resulting insights to predict events that might occur on the network. 
 

Studying the data to prevent outages, fight cybercrime and improve customer service 

We can break these events into three categories: 

• Network faults and anomalies 

  Network outages annoy customers, which damages an MNO's reputation and costs money. Why? Because unreliable service is one of the biggest drivers of churn. It also leads to employee overtime and possibly penalties for not meeting service level agreements (SLAs). 

  In the 5G world, networks will generate millions of alarms that signal unusual behaviour. MNOs can use analytics to learn which of these alarms are most likely to lead to a fault. This is known as predictive maintenance. 
   

• Cyber attacks 

  Cybercrime is rising all the time. The 2021 SonicWall Cyber Threat Report reported 268,362 'never-before-seen' malware variants in 2020. It also referenced 56.9 million IoT malware attempts (66% up on 2019).

  One key reason for the rise is the availability of new artificially intelligent tools – AI-based malware that can enter a system without detection and subtly change it from the inside. Attackers can buy these off-the-shelf products on the dark web. 

  In the new 5g data environment, existing rules-based detection systems (that look for well-known threats) won't suffice. Instead, MNOs must employ advanced analytics tools to detect more subtle patterns. 
   

• Changing customer behaviors 

  The explosion of 5G data gives MNOs an unprecedented opportunity to understand individual subscribers' behavior. They can give marketers access to data consumption patterns, web browsing habits, mobile application use, location data, etc.  

  Armed with these insights, marketers can micro-segment their targets. They can devise effective promotions, ad campaigns and more. The end result is a better customer experience, reduced churn and more revenue. 

Complicating factors: private network slices and the IoT 

Needless to say, the sheer volume of 5G network traffic will make it challenging for MNOs to analyze 5G data effectively.  

But it's not just scale that's a problem. There is also the issue of how the new networks are constructed and who 'owns' the data. 

The first complicating factor is network slicing. The virtual nature of 5G infrastructure makes it possible to create smaller network segments at the 'edge'. MNOs can sell these 'slices' to private companies, who can customize them as required. 

However, these private networks will vary in their degree of isolation. Some will be independent, on-premises deployments with limited interoperability with public networks. Some will sit on top of infrastructure that is shared with public services. Other 'hybrid' private 5G networks will combine public infrastructure with infrastructure at a customer's premises. 

Each one will change how data flows across the network, how much automation is involved, and who can access it. 

Then there's the IoT. As we have suggested, 5 G's huge capacity and range will make connecting billions of machines and sensors possible.  

But clearly, 'things' are not like humans. Often, they send data in short bursts and over longer periods. They cannot respond to alerts as people do. 

These differences and others compel MNOs to find a new way to analyze the data moving over connected devices. 
 

An analytics standard specially designed for 5G networks: NWDAF 

Happily, when the 3GPP standards body worked on next-gen networks, it devised a standard to build machine intelligence into the 5G core: NWDAF (network data analytics function). 

NWDAF represents the mobile industry's first attempt to standardize analytics function in the mobile core network.

With NWDAF, the 3GPP ensures that analytics are not an afterthought and that analytics-driven automation is built into the system architecture.
 

Here come Thales' Guavus analytics solutions 

To help MNOs navigate the complexity of 5G data analytics,  

Thales has developed a series of solutions (collectively named Guavus). 

Telcos can deploy Guavus tools for two purposes. The first is to keep the network running optimally. 

These tools are collectively named Ops IQ. MNOs can configure them to set up millions of alarms, which are graded so that engineers can stay focused on critical issues or incidents affecting subscribers' quality of experience. 

With the application of machine learning, these systems can deliver predictive maintenance – the ability to anticipate faults before they occur.

This approach has saved CSPs upwards of $10 million a year in OPEX.  

The Ops-IQ suite consists of three modules: 

• Network fault analytics 
  Advanced alarm management for network and service operations centers. 
• Service Experience  
  Next-gen proactive assurance for network, service  and care operations teams 
• LiveOps 
  Real-time operational insights for network, field  and customer care teams 

The second group of Guavus services were developed for customer experience. They are called Service IQ, allowing marketers to understand individual subscribers' behavior by looking at digital interactions with the network.  

Service IQ products consist of two modules: 

• Marketing Analytics  
  Subscriber behavioral analytics for marketing operations and product teams 
• Device management analytics 
  Analytics for network planning, product and marketing, and network operations teams.