AI and edge computing trialled to reduce telco bill strain
As 5G increases data usage and new multi-partner services proliferate, billing will become more complex and resource-hungry. China Unicom and a team of partners are trialling a new approach to address the issue.
29 Nov 2019
AI and edge computing trialled to reduce telco bill strain
As 5G increases data usage and new multi-partner services proliferate, billing will become more complex and resource-hungry. China Unicom and a team of partners are trialling a new approach to address the issue.
In 2018, average monthly smartphone data consumption stood at 5.6GB per month globally. By 2024, as 5G rolls out, this is expected to at least quadruple to 21GB per month, driven by the increasing use of video and new applications such as augmented and virtual reality. The growth in smart buildings, cloud services, smart cities, industrial automation and more will further increase data usage exponentially.
This data explosion is demonstrated on perhaps the biggest scale in China. In 2018, the number of mobile internet users in the country reached 793.62 million, representing 97.59% of total internet subscribers. There are expected to be 1.18 billion internet subscribers by 2027. That’s a huge amount of data for telcos to manage.
Billing will become more complex as new 5G-enabled services require on-demand provisioning, real-time service creation, order management, dynamic online charging, service level agreement (SLA) creation and monitoring, and more.
All this will need to be processed and matched with packages in real-time, which will place new pressure on network operators’ support systems, and will require additional bandwidth, equipment, computing power, service capabilities and storage resources.
These challenges are being addressed by a TM Forum proof-of-concept Catalyst project, titled AI for 5G billing data acceleration.
China Unicom is the Champion of the project, setting the business challenge and Si-Tech, Whale Cloud and BUPT are participants, who will collaborate to deliver different aspects of a proposed solution over a short three-month sprint.
Sun Mingli, Si-Tech, said: “The problems of storage, bandwidth and system are urgent to solve.”
For example, in one province in China, China Unicom noted that between September 2018 and July 2019, the number of credit control request (CCR) messages across its network increased by 126%, meaning storage needs increased by 263%. On average, 1.57 billion CCR messages were generated each day in July 2019, contributing to a large volume of data daily.
To help China Unicom and other telcos manage this, the team is developing a platform which uses artificial intelligence (AI) and machine learning, as well as edge processing.
Conventional strategies use a fixed data quota policy for each user. The Catalyst team is proposing a platform to enable a more dynamic scheme which takes into account various factors to become more predictive.
Based on user tags and usage record information, the ‘AI neural network intelligent quota prediction model’ generates’ predicted data by users, services and time and assigns an ‘intelligent data quota’, which can be matched against actual usage in real-time.
This predictive capacity reduces CCR messages and call detail records (CDRs), and therefore minimises strain on support systems.
When the system was trialled in a province of China by China Unicom, the volume of CCR and CDR messages decreased by 63%.
To reduce strain further, the team is also exploring the use of compression techniques and a gateway that allows billing messages to be processed at the edge rather than in the core network.
At Digital Transformation Asia in Kuala Lumpur (November 12-14), the team will demonstrate the improvements their platform can deliver compared to traditional systems.
The platform has been developed using TM Forum Open APIs, and the team will provide their findings back to the TM Forum community so the tools and best practices can be extended.
Sun Mingli said: “China is a vast country with a large population. If our platform can solve this challenge in China, we can solve it anywhere so it is very relevant to other operators facing the same issues.”
In 2018, average monthly smartphone data consumption stood at 5.6GB per month globally. By 2024, as 5G rolls out, this is expected to at least quadruple to 21GB per month, driven by the increasing use of video and new applications such as augmented and virtual reality. The growth in smart buildings, cloud services, smart cities, industrial automation and more will further increase data usage exponentially.
This data explosion is demonstrated on perhaps the biggest scale in China. In 2018, the number of mobile internet users in the country reached 793.62 million, representing 97.59% of total internet subscribers. There are expected to be 1.18 billion internet subscribers by 2027. That’s a huge amount of data for telcos to manage.
Billing will become more complex as new 5G-enabled services require on-demand provisioning, real-time service creation, order management, dynamic online charging, service level agreement (SLA) creation and monitoring, and more.
New pressures
All this will need to be processed and matched with packages in real-time, which will place new pressure on network operators’ support systems, and will require additional bandwidth, equipment, computing power, service capabilities and storage resources.
These challenges are being addressed by a TM Forum proof-of-concept Catalyst project, titled AI for 5G billing data acceleration.
China Unicom is the Champion of the project, setting the business challenge and Si-Tech, Whale Cloud and BUPT are participants, who will collaborate to deliver different aspects of a proposed solution over a short three-month sprint.
Sun Mingli, Si-Tech, said: “The problems of storage, bandwidth and system are urgent to solve.”
For example, in one province in China, China Unicom noted that between September 2018 and July 2019, the number of credit control request (CCR) messages across its network increased by 126%, meaning storage needs increased by 263%. On average, 1.57 billion CCR messages were generated each day in July 2019, contributing to a large volume of data daily.
To help China Unicom and other telcos manage this, the team is developing a platform which uses artificial intelligence (AI) and machine learning, as well as edge processing.
Becoming predictive
Conventional strategies use a fixed data quota policy for each user. The Catalyst team is proposing a platform to enable a more dynamic scheme which takes into account various factors to become more predictive.
Based on user tags and usage record information, the ‘AI neural network intelligent quota prediction model’ generates’ predicted data by users, services and time and assigns an ‘intelligent data quota’, which can be matched against actual usage in real-time.
This predictive capacity reduces CCR messages and call detail records (CDRs), and therefore minimises strain on support systems.
When the system was trialled in a province of China by China Unicom, the volume of CCR and CDR messages decreased by 63%.
To reduce strain further, the team is also exploring the use of compression techniques and a gateway that allows billing messages to be processed at the edge rather than in the core network.
At Digital Transformation Asia in Kuala Lumpur (November 12-14), the team will demonstrate the improvements their platform can deliver compared to traditional systems.
The platform has been developed using TM Forum Open APIs, and the team will provide their findings back to the TM Forum community so the tools and best practices can be extended.
Sun Mingli said: “China is a vast country with a large population. If our platform can solve this challenge in China, we can solve it anywhere so it is very relevant to other operators facing the same issues.”