5G industry news

How AI is handling 5G data surges

  • 5 minute read
  • Published by Arti Mehta on 16 Jun 2020
  • Last modified 16 Jun 2020
Since the launch of 4G, operators worldwide have seen explosive growth in traffic as data limits have risen, prices have fallen, and the networks’ capacity and speed increased.

A version of this article was originally published on TM Forum's Inform information hub.

This was China Unicom’s experience, accompanied by soaring number of messages concerning credit control requests (CCRs), generated by the Diameter Credit-Control application protocol which defines a charging mechanism for pre-paid users. To do this, it uses a credit-limit control to implement session- and event-based charging. Likewise, the rise in traffic also produces huge numbers of call detail records (CDRs).

Some idea of the scale of the problem is shown by the fact that in Shandong province alone, China Unicom found the number of CCR messages rose by 126% in less than a year, with the average daily volume of them reaching 1.57 billion. This put immense strain on network elements, storage and billing, and the many systems from which billing takes information.

Traditional billing solutions cannot handle the problems caused by this steep rise in 4G traffic as they are typically designed to deal with a fixed quota, such as 30MB for 30 minutes. In addition, the rapid advance of 5G services brings an unprecedented surge in traffic and message volume, putting huge pressure on computing and storage resources. More investment and energy consumption are needed.

Taking the strain

There is an urgent need to reduce the strain on and soaring costs of storage, bandwidth and systems associated with billing. At the same time, operators need to improve operational efficiency as well as to implement global governance strategies and invest heavily in new hardware and especially software technologies – and all with without negatively affecting users’ experience.

Against this backdrop, China Unicom decided to explore how artificial intelligence (AI) could be used to address these challenges. In parallel it wanted to investigate how edge computing could reduce the pressure on the network elements across the infrastructure and their associated systems, as well as help with the expected burst of traffic at the edge.

Using AI to build models

As a first step, the operator used big data analysis to determine the most effective way to train the AI. The AI monitors how the models run, then ‘learns’ from outcomes, and through continuous iteration of the models optimizes their performance.

From the tags, the models predict data about users, services and time periods, and assigns an intelligent data quota, which can be matched against actual usage in real-time. This predictive capability reduced CCR-related messages by 63% in Shandong province, which equates to 964 million fewer CCR messages every day (also see results below).

By the end of the trial, China Unicom had cut, by 50%, the bandwidth and computing resources needed to handle the billing and CDRs efficiently. This massively reduced the amount of processing and response delays from network elements, the online charging systems and the business service inquiry system, such as for balance and account inquiries, which is related to 5G billing, which were not part of the trial. All of which also meant that customers noticed that the service had improved.

Other systems that are connected to the 5G billing system, like big data platforms, will also save a lot of storage space and improve the overall efficiency of the systems’ processing.

AI is the key

While the data analytics are used to mine the data held in core billing systems, and to identify that value through business operations, it is the AI which realizes that value. The AI-powered intelligent billing in this project makes a strong use case for operators to employ AI technology to empower their cloud-based systems.

This project also covered edge computing. By dynamically allocating traffic quotas based on users’ online behaviour, the amount of data is reduced significantly. This means that after data processing at the edge, only the useful data is sent to the billing system.

In future, new use cases will be explored such as: automatic switching of online and offline charging based on the user’s credit, balance and other information; and intelligent control based on the user’s 5G context, such as low-latency scenarios. China Unicom will also consider putting part of its computing capabilities on the edge of the network.

Results

  • Daily CCR messages reduced by 964 million equating to a 63% fall from 1.57 billion, which reduced the monthly total of CCRs to 28.917 billion (30 days x 796 million).
  • The daily number of CDRs fell by 52%, to 468 million from 900 million, which reduced the monthly total to 14.040 billion (30 days x 468 million)
  • Average amount of storage needed daily before going online was 3.86TB, which fell by 1.84TB when operations moved to the containerized platform
  • This reduced monthly storage by 55.2TB (30 days x 1.84TB)
  • The system stores 13 months’ of data, hence over that period the amount of storage required has fallen by 717.7TB (13 months x 55.2TB)

As China is the world’s third largest country in terms of landmass and has the largest population, the success of this project suggests if this platform and approach can solve the challenge in China, then they can be implemented anywhere on the planet by operators to address the same issues.

This article was provided by a contributing writer to TM Forum.

TM Forum is a global industry association for service providers and their suppliers in the telecommunications industry.

Tweet