Page 358 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

2 Use Case Description

2�1 Description

In the context of the global energy crisis, many parts of the world are struggling with power
supply issues, and internet access is frequently disrupted by power outages and restrictions,
exacerbating the digital divide in these areas. For example, in South Africa, rotating power
outages can sometimes last up to 10 hours [1], causing base stations and routers to shut down
and available links to decrease, resulting in network congestion and packet loss that negatively
impact user experience. Analyzing individual base stations in manual O&M (Operations and
Maintenance) mode requires combining a significant amount of data. However, traditional
coding methods are inefficient and inaccurate for data fitting analysis.

AI capabilities of the telecom foundation model and the network data from the digital map are
used in the solution to enable the automation of the following network analysis and optimization
tasks:

Automatic traffic suppression analysis and real-time visualization:

Based on the network digital map and the network optimization agent's built-in algorithm, the
system can collect traffic data at a millisecond level, providing a unique heat map to accurately
restore the suppressed traffic distribution of base stations and visually identify network
congestion points in real-time. The algorithm models the relationship between packet loss
rate and traffic rate, allowing it to calculate the exact amount of suppressed traffic.

Precise global path optimization:

Before any optimization can be effectively applied, it is essential to have network awareness —
a real-time understanding of the network topology, link status, bandwidth utilization, latency,
and potential congestion points. By continuously sensing and responding to dynamic network
changes, the system ensures that optimization decisions are made based on the most current
network state. Based on graph theory [5] and operations research, a comprehensive algorithmic
framework is designed to support a flexible combination of diverse computing policies. And
leveraging SRv6 technology, the system enables fine-grained and programmable path control
across the IP transport network, allowing services to be steered along optimal paths that meet
their specific SLA requirements in various scenarios. This approach not only improves service
reliability and performance but also enhances network resource utilization and operational
agility.

Predictive operations:

a) Provide minute-level prediction of suppressed traffic through a characterized time-
series model for base stations, offering precise capacity expansion suggestions for traffic
growth, such as the World Cup.
b) Setting network SLAs in advance, if the packet loss rates or delays of network services
reach the warning thresholds due to high link bandwidth utilization or power restrictions,
our system will automatically design network paths to enhance service quality. The system
offers more than 15 path calculation factors, including bandwidth, latency, and availability,
among others. These factors can be flexibly combined to calculate the optimal path. This
helps operators classify services more precisely and route them to network paths with
different Service Level Agreements (SLAs). As a result, both service-level agreements
(SLAs) are met, and resource utilization is maximized.

322